Assembly Standing Committee on Privacy and Consumer Protection

On for all morning. Yesterday, I was with one other person.

Good afternoon. Got a corner.

Welcome to Assembly privacy and consumer protection. We are so popular today, so we're glad you're all here. We will be missing Assembly member Wicks today, but we welcome Assembly member Bennett in her place. We have 17 bills on the agenda today. One on the consent calendar, which is a Senator Umberg bill, SB 96, and, we will take a motion on that once we have a quorum.

Members, feel free to stop by so we can get that quorum going. To effectively manage our time today, we'll be litigating a testimony to two witnesses in support and two in opposition, two minutes for everybody presenting. It's required that primary witnesses have communicated with the committee staff and submitted to the committee their supporter opposition prior to today's hearing. After the support witnesses, we will make time for everybody to give us their name, organization, and position.

If you do not have time to give us everything you wanna say, please use our portal online.

We love to see, people's comments, and we do read those letters and process those thoughts, and that is still available to people even after today's hearing. With that, don't disturb people. I keep decorum, and we will We're gonna we're just gonna go for it, Senator Umberg. So we're gonna take a SB 574, Mister Senator Umberg's bill when you're ready, Senator.

Thank you very much. Thank you to the hard work of the committee, particularly Mr. Josh Tosney, for your work on this bill. And you're welcome too, Mr. Tosney, for the tutelage you received in Senate Judiciary. Senate Bill 574 protects clients receiving legal services by establishing rules for the use of AI by attorneys, judges, and court neutrals. Artificial intelligence, of course, is present in all aspects of our life, including in the law. What this bill does, it just adds transparency to the use of artificial intelligence.

Transparency when lawyers submit materials, transparency when judges use artificial intelligence to arrive at a decision, transparency in terms of arbitration so that there's actually a human who makes all final decisions. Clearly, artificial intelligence has a very useful use in the legal practice.

But it cannot supplant human beings and cannot supplant judgment and certainly cannot be such that you can blame artificial intelligence or any model for an ultimate decision or an ultimate submission. If you've suggested amendments, I'm sure they were really good, and I accept them. So.

Thank you. Do you have a witness?

I'm it.

You're it. We'll take it.

Do I need more?

You never need more, Mr. Umberg. You are enough. So with that, we will take any other individuals here in support of this bill, SB 574. Come on up. Name, organization, and position. Tracy is always here.

Yeah. Good afternoon, Chair and Members. Tracy Rosenberg, Oakland Privacy, in support of the bill.

Anyone here in opposition to this bill? Seeing none. I'll bring it back to the Committee Member. Assembly Member Irwin.

I feel like I do have to ask a few questions. You can't get off this easy. Well, I think it is really important to start to put some guidelines around this. It happened. I was talking to a girlfriend of mine who was talking about, she's an attorney. And she was talking about the AI hallucinations and these court cases that were cited that were not actually court cases.

So I think that that is, you know, it's really important that we start to deal with that. But I, yeah. And then I also ran it by my daughter who said, great idea, except she was wondering where it says attorneys must, from the documents, remove bias or harmful AI generated content. So how do we define that? It it might be a little nebulous.

Well, I suppose that that's in the eye of the beholder. But if artificial intelligence can generate obvious errors, can generate obvious bias, And so when a lawyer sees that, when a lawyer sees obvious bias, they have to eliminate. They also have to make sure that they don't submit personal information. So if you're putting somebody's Social Security number in a complaint, you need to file under seal. If it's not necessary, you shouldn't be putting in a complaint.

That kind of thing. So attorney's responsible to make sure that whatever document they submit because everything, as we know, is subject to being scraped from the web. So you don't want, you don't want documents, pleadings, etcetera, filed that may be used for nefarious purposes, inappropriate nefarious purposes.

Or you could require that they have enterprise systems, right, where all the data stays in the model and it's not being used to train the model for other people's purposes.

You could, but virtually everything that's filed is accessible to the public and has to be accessible to the public. And so unless you file it under seal, it's anybody, can anybody can view it and access it. So.

Well, I would be happy to move the bill at the appropriate time and will certainly be supporting it.

Thank you very much.

Thank you, Assembly Member. May I note that I love that Assembly Member Irwin's daughter is now her lawyer advisor. It's amazing. So thank you, Senator, for this bill. I think it's incredibly important, as a fellow member of the Bar. I've gotten the emails from the Bar where they're advising us on how to appropriately use generative AI in the practice of law.

And I was recently with some of my former colleagues who are still in practice asking them about their use of it. And I think that I was comforted by the conversation because I think lawyers are well trained to understand our duty and our obligations under our license. And so they are actively using it, but they very much are keeping the human in the loop.

So they consider the AI to be draft one, and then it is their obligation to review the draft, to take out anything that has apparent bias, to check any citations if there are any or anything of that nature. And I think that it's important that we as attorneys, when we're hired, we're hired because people trust us.

And that trust is critically important. And outsourcing it without the kind of oversight that this bill would require would, I think, be a dereliction duty. And so I think this is a great first step, but I know that the Bar will continue to do additional work as well, as the bill requires. So I'm happy to support it today. Would you like to close, Senator?

No. I'm pleased that you're both hanging out with very responsible lawyers. But all lawyers are not as responsible as the the folks that you interact with. And so this is to, in essence, put them on notice and to the extent the courts need notice as well that you have to be transparent. I've in my own practice, I've seen hallucinations. I've seen people who have submitted materials that not long ago, I saw a citation to a code section that doesn't exist in California. So it's appropriate that we put this in law.

Yeah. Well, thank you. When the time's right, we'll take a motion. And thank you, Senator, for this and the many years of bills you brought to us. We will miss you.

Well, okay. The reports of my demise are a little early.

Not in the Assembly. This is the last time you're speaking before me for sure in the Assembly.

Oh, no. I'm sure we'll cause trouble in August. So.

Okay. Fair enough. Then we'll see you then.

Thank you very much. Okay. Thank you.

Okay. So we're looking for some Senators and, frankly, some Assembly Members here in 444 for the Assembly Privacy Committee. Senator Padilla would be next on our agenda, but if any Senators are wanting to get out of here on the earlier side, this is your opportunity to hop on by.

Okay. We have an author because Assembly member Pellerin will be presenting Senator Rubio's bill, SB 1276. So we will hear expert here. This time.

I'm really excited. Okay. So it is my great honor to be here today to present, SB 1276, the End Child Exploitation Act on behalf of Senator Rubio. Emerging technologies such as AI and live streaming have enabled a new class of criminals to access and proliferate child sexual abuse content, presenting new and unique challenges for law enforcement who often struggle to keep pace with the new technologies. Existing law criminalizes the creation and distribution of live streams depicting the sexual or exploitation of a child.

However, there is ambiguity in how those who view the content on live streams are treated under the law. Platforms like Facebook Live, TikTok Live, Discord, and Twitch enable temporary streaming in which the content disappears after the stream ends, leaving little record of the content of the video itself. Child predators can take advantage of this feature by consuming child exploitation content and exiting the stream, leaving no trace of their crime.

SB 1276 makes a person guilty of a misdemeanor or felony if the person downloads, streams, or accesses through electronic or digital media depictions of a minor engaged in a sexual act. The bill also adds digital and AI generated images to the definition of sexual exploitation under the Child Abuse and Neglect Reporting Act, clarifying that it is a crime to watch a livestream and AI generated content of a sexually exploited child.

As technology develops, there should be no loophole available to sex track to child sex predators. By removing ambiguity in the law and updating it to include modern technological standards, SB 1276 protects our most innocent, that is our children. Joining me today to speak in support of twelve seventy six is Laura a Drino, the director of Crimes Against Children.

Thank you. Good afternoon, everyone. Again, my name is Laura Drino, and I am, with the Los Angeles City Attorney's Office and handle everything involving crimes against children. I'm here in strong support of SB 1276. I've been a prosecutor for almost thirty three years now, and this bill closes a really a a real gap in the law by harmonizing the law of mandatory reporting under a camera and penal code section three eleven point three child sexual exploitation.

As a prosecutor who's tried many of these types of cases over my career, predators do not stop adapting. They use new technology to exploit children. They use new platforms to exploit children. And increasingly, technology has changed the way that offenders access and consume this type of material. They now access child sexual abuse material through live streams and digital media rather than traditional downloads.

And we've seen a lot of cases growing in Los Angeles in this area, especially on some of the newer apps like Discord. We've had many cases. Our law must reflect that reality. A child being sexually abused on livestream is no less a victim because the abuse is viewed in real time instead of from some downloaded or saved file. The trauma is the same for that child.

The exploitation is the same, and the demand created by every viewer is the same. Twelve seventy six makes it very clear that knowingly watching a livestream of a minor engaged in sexual conduct is sexually exploiting that child under penal code section three eleven point three. Also important is to harmonize three eleven point three with the Child Abuse Reporting Act. The definition of sexual exploitation needs to be consistent for prosecutors, for law enforcement, social workers, teachers, and mandated reporters across the state.

No offender should escape accountability because technology moved faster than our statutes.

The opposition is saying we need to focus our criminal laws on behavior, which harms actual victims and not on severe punishments based on whether a person has a deviant sexual interest. This bill is focused on actual victims. Children's photos are being taken off in in addressing the AI issue. Children's photos are being taken off public social media sites. Many parents or people are public on Instagram, on Facebook.

They post pictures of their family, of their children. Those photos can be taken off of real children and then used to generate AI images and video and livestream of the the face of this child engaging in sexual conduct. There have been cases where there's AI generated conduct of a young girl orally copulating an adult male. That's a real child. That's that child's face that is now out there for people to view for the rest of their life.

So the opposition saying that it doesn't harm a real child, it does. The trauma from that will last for that child's lifetime into adulthood. So let's not excuse this as a sexual deviance. This is child sexual exploitation of a child. California must be clear.

The child sex sexual exploitation in any form streamed digital AI generated will be reported and prosecuted, and I respectfully ask for your aye vote. Thank you.

Thank you so much for your testimony. Anyone else here in support of this bill? Name, organization, and position, please.

Hi there. Dawn Koepke on behalf of the Child Abuse Prevention Center in support.

Thank you.

Good afternoon. Olivia Herrera, intern at Stone Advocacy on behalf of the California District Attorneys Association in support. Thank you.

Thank you. Not seeing anyone anyone else here in support? K. I don't see any Is the opposition here?

You mean Actually, I'm in support. I'm currently in support with the Latin American and Caribbean Business Chamber of Commerce and the Hispanic Chamber of Commerce San Francisco. Support because we know how it is happening with all the children that come in and cross over here. Thank you.

Thank you so much. Is the opposition here? I know we have opposition on file. Okay. Well, the opposition didn't show up.

So anyone else here wanna register their opposition to this measure? K. Thank God the answer is no.

Senator Irwin. Yes. And and thank you very much for working to close this loophole, and I completely disagree with the opposition that there is no harm. So, appreciate you bringing the bill forward for, Senator Rubio. I was wondering.

I I thought that there was, that there were technical issues with you know, obviously, when something is downloaded, it's easy to prove that it's in possession.

Right.

So are there technical issues with trying to prove that somebody was streaming child pornography or AI generated child pornography?

Well, it's a it's a different level, but law enforcement is able to able to to enforce that. There's the whole knowing requirement whether the person knew that they were doing it, and that is addressed in Subdivision G Of 311.3. So it is able to be proved as far as whether somebody's viewing livestream.

We've had cases recently on the Discord app where a group of young men in another state and some in California are observing sexual conduct, and law enforcement's able to do search warrants and find out who has been, live streaming these children. Okay.

And and I also wanna say the work you do is so extremely difficult, but it's so incredibly important, the the lives that you are basically saving. So thank you for that.

Thank you. I'm very passionate about it, so thank you.

Thank you, Assemblymember Erwin. And I just wanna express my gratitude too. It must be really hard to do this every day and see these cases, and but we need people to do it. So I really appreciate you taking this on for the people of Los Angeles, but probably to the point of this bill far beyond because a lot of this is happening online. So I understand, and I appreciate the answer.

I'm assuming that much of this is, you know, if it's on Discord, you serve a word on Discord who was viewing this, and they have records, and that's how you're able to prove Right. What was going on. Can I ask, is it criminal behavior for the platforms to be facilitating?

The more difficult for the platforms, it's more difficult for them to view it in real time, the live streaming, to take

it down. But I know that that's something some platforms are, you know, working on. But it is it's easier for the platforms to see when people are downloading or doing different types of activities than it is the actual live streaming. But, you know, that's kind of beyond beyond my mind.

Right. No. I was just the devices already have artificial intelligence, I know I think Amazon provides one that people use to screen images. I don't see why AI couldn't be an answer that would allow us to screen and stop these streams in order to prevent some of the cross paths. Again, not again, not instantaneous, but with some minimal delay.

I just think that that's something that should be incumbent upon the platforms as well so that we're not having this happen. So it just it's horrendous, and I just I I was at the UN Internet Technology Conference and had the opportunity to speak with some legislators from The Philippines where a lot of this content is created and a lot of their children are being abused.

Right. I

And it is just I think we need to do everything we can to partner with people even across the globe to ensure that kids are not being harmed in this way. So just curious what is what we could do.

Unfortunately, we've had two cases of children right here in Los Angeles that are being sexually exploited and livestreamed and told to after the sexual conduct to harm themselves, commit suicide, all of those things. So it's it's really dark.

It is really dark. We need the promises of technology without the perils, and this is really the perils laid bare. So thank you for that. This is obviously an incredibly important bill that I am eager to support today. Assembly member, do you wanna close on behalf of Senator Rubio?

Yes. Well, the good news is she reported that the bills received bipartisan support. And on behalf of Senator Rubio, I respectfully ask for your aye vote. Thank you.

Thank you. When the time is right, we'll take the motion and the vote. Oh, we have two senators, I think. I think Senator McNearney is standing behind sitting behind you. So Mister Mcner Senator Mcnerdy, you are up first with SB 813. Yeah.

When you are ready, Senator, we are ready for you.

Well, thank you, Chair Bauer-Kahan. Now I'm glad you got the memo on the sports jacket. I'm here today to present SB 813. I wanna start by thanking the Chair and the committee staff for their amazing work. There's been a lot of back and forth on this, and I think it's pretty close to being what we would find ideal here.

I do accept the committee amendments. I wanna thank all my colleagues who have taken time to speak with me about this proposal as well. I also wanna thank the stakeholders who have engaged in the discussions. SB 813 is a critical step in making AI safe and beneficial. SB 813 would establish a commission to oversee committees that will create safety standards for AI development and use cases.

I've worked on standard spot as before and I know the importance of having a robust process that includes industry, labor, academics, and consumer advocates for creating proactive standards for emerging technologies to prevent harm before it occurs and learning from past harms to design safety into otherwise potentially dangerous technology. The bill embodies the regular standards making process as seen in many other national and international standards bodies such as NIST, UL, or ISO. So, why do we need standards for AI safety in California?

It's because California cannot wait for Washington. AI capabilities are advancing faster than traditional regulatory processes.

California cannot wait for years for a paralyzed contracts to act. The states, including California, are beginning to establish guardrails for AI to protect families, to protect workers, to protect the security of our elections and, of course, to protect our children. But these standards could improve could unravel if the federal government did enact legislation to prevent all state AI laws and there's been motion to do that.

SB 813 gives California the opportunity to fight back against federal preemption of state AI laws by establishing a framework that is not a regulation. It's a voluntary process.

SB 813 would establish a two tiered framework for AI safety standards, a minimum compliance standard, and an advanced safety standard. The commissioner will then - the commission will then convene stakeholders, working groups of standards bodies to solicit input from the development and revision of these standards. It's important to note that SB 813 does not work alone. I have a partner in crime here working with me, sitting in the Chair there.

SB 813 works together with the Chair chair's own bill, AB 145. AB 145 creates a state registry of AI auditors, who use these standards that are created by SB 813 when performing third party assessments to evaluate AI systems or models. This partnership creates a first in the nation, a scalable AI government framework. And by scalable, I mean that we can scale this to other states, to the nation, and maybe to other countries around the world.

It's important that we do something now.

The lack of action has created increasing risk and great urgency. We cannot wait for AI to control us. We must remain in control ourselves. Here and to testify with me today is Nicole Rocha of Fathom and Jai Jaisimha from Transparency Coalition AI. Jasmine?

Two minutes each when you're ready.

Good afternoon, Chair and members. I am Nicole Rocha here on behalf of Fathom, the sponsor of 813. Fathom is an independent, nonpartisan, nonprofit whose mission is to find, build, and scale policy and technical innovations in the public interest. Fathom has spent the last two years developing and championing a practical, adaptable framework for AI governance built on a simple and powerful premise. Government sets public interest goals and independent experts verify results.

SB 813 creates a two tiered system by which by which both goals are accomplished. It establishes the California artificial standards and safety commission, which in turn designates and oversees a marketplace of place of auditors and IVOs tasked with evaluating compliance.

Under this structure, the commission sets desired public interest outcomes, including product safety, child well-being, and data security among others, while independent expert led auditors and IVOs develop the technical criteria to determine whether AI models meet those outcomes and then verify AI systems that voluntarily seek certification against that criteria. We have seen, in other context, how voluntary frameworks like the one envisioned by 813 create a race to the top, like with environmental standards.

The success of this framework largely depends on the independence of the professionals conducting the assessments, And 813 demands that auditors and IVOs are free from conflicts of interest.

Independent third party evaluation offers decisive advantages over traditional regulatory models. First, it moves at the speed of innovation. Second, it elevates expertise. And third, it is designed to scale. Just as professional licenses and standards bodies operate across state lines while record recognizing each other's credentials, the independent third party auditors and IVOs can create a facto national standard in the absence of federal action, ensuring that California will continue leading the nation on technology governance. Thank you for your consideration, and I'm happy to answer any questions.

We ask for your aye vote.

Thank you.

Honorable Chair and members of the Assembly Privacy Committee. My name is Jai Jaisimha, and I'm proud to testify today on behalf of the Transparency Coalition in support of SB 813. While industry critics - I'll talk a little bit about some of the opposition arguments and and how we feel about it. While industry critics might argue that new compliance tiers will slow down development and stifle California's competitive edge, the reality is exactly the opposite. Innovation requires public trust and market predictability to thrive.

SB 813 provides a stable foundation. Furthermore, the bill explicitly prevents regulatory stagnation by utilizing structured working groups. These groups ensure that industry developers sit at the table with academics and civil society to co-create evidence based standards that evolve with the speed of tech. I used to I've been part of a similar process in the state of Washington with stakeholders from all aspects of society that actually ended up producing much more inclusive and protective policies.

We've also heard concerns about this set of standards being state specific as opposed to relying on international standards.

While the NIST RMF, the risk management framework, has been in place for some time now, all we've seen from industry is a lack of action. Industry continues to express support for existing voluntary commitments, but failed to demonstrate their actual compliance with any transparency, and the only constant is federal inaction. The provisions outlined in this bill and AB 145 collectively create a best in class audit framework that also builds on decades of industry practice in other areas.

As someone who has built or complied with audit and certification programs at both startups and large companies like Microsoft, I can state that they are completely feasible. All it needs is a little encouragement from the California state legislature.

By establishing the California artificial intelligence standard and safety commission, SB 813 guarantees our basic consumer, worker, and business protections are met. Thank you so much for giving me the opportunity to testify, and I respectfully ask for your aye vote.

Thank you. Anybody else here in support of this measure? Seeing no one. Do we have primary opposition on this measure? I know we had registered.

Yep. Come on up.

Good afternoon.

Good afternoon, Chair members. Robert Boykin with TechNet. Here today, to testify in respectful opposition to a to SB 813. And I'll say, my comments today onto the bill in print, because I know there has been amendments and that agreed to. Right?

So but let me start where we agree. Standards and independent audit? Yeah. Okay. There's our amendments in the in the analysis.

Right?

Yeah.

Okay. Cool.

Let me start to where we agree. Sanders independent auditing can play a real role in AI safety. We appreciate the author's continued engagement and the restructuring in the June 16 amendments, the two tier framework, and a new stakeholder working group. But as amended, the bill still asks seven member state commission to do something that the body simply is in position to do.

Setting advanced AI safety standards takes deep current expertise in model evaluation, red teaming, and security, the kind that lives at NIST, ISO, and IEEE bodies with decades of consistence based experience and thousands to and thousands of contributing experts.

SB 813's commission would be seven political appointees, none required to have standard setting experience, as they evaluate hundreds of existing standards and write new ones. Core terms like heightened care, best practices, substantially exceed minimum compliance, remain undefined with no required public comment, peer review, or consensus approach. A working group is a welcome step, but it's not a rigorous standards process. NIST's AI risk management framework took years of public comment and peer review. This bill requires none of that.

The framework is also voluntary in name only. Earlier versions carry the liability shield that gave developers a reason to certify. That's gone. And what remains is the market pressure from insurers, procurement, and plaintiff's attorneys, that pushes companies to certify anyway. That's the de facto mandate without the legal certainty of a real incentive we provide.

And in fragments, this committee is concurrently negotiating audit and assessment requirements in SB 1119, AB 2023. Standing up a separate two tier auditing regimen under SB 813 risk two parallel potentially inconsistent frameworks for the same technology on top of a growing patchwork in Ohio, Connecticut, and beyond. California has led by aligning with broad globally recognized standards, not by building siloed ones other jurisdictions won't recognize. Also, we'd urge the same approach here.

For these reasons, we respectfully oppose eight one three and ask for a no vote today.

Thank you.

Thank you.

Missoni? Thank you. Ronak Delami on behalf of Cal Chamber in an imposed and less amended position. While let while our letter goes into greater detail, there are a few issues we also wanna highlight for today's hearing. Cal Chamber does share the goal of promoting safe and trustworthy AI, but we are concerned that this bill creates a California specific development and auditing framework at a time when robust national and international standards are already being developed as mentioned by my colleague.

Technical standards should be developed through expert driven consensus based processes, not through a structure that risk becoming outdated or inconsistent with broader national and global efforts. We're, of course, reviewing the committee amendments but, on our first three, the amendments failed to address the core concerns. Our core concerns including those around the breadth of the authority and expertise of the commission or the significant resource and capacity constraints involved in reviewing and developing auditing protocols for rapidly evolving technical standards.

Realistically, international bodies rely on thousands of experts over multiple years to produce a single standard. So without the proper resources to do this, we're simply going to create an immense regulatory bottleneck and risk rendering state guidelines obsolete before they are even published.

We also point out that, things such as, trying to align and aligning are two different things. Again, while we share the author's goal of advancing AI safety while preserving innovation and interoperability, we do have significant reservations over this approach at this time. But some of the significant amendments are done to address these and other concerns in our letter. We do respectfully ask for a no vote at this time. Thank you.

Thank you.

Anybody else here in opposition to this measure? Must be really busy around here. I just have people here where nobody's coming up for anything.

Tracy was with with open privacy. We were in opposition to the previous version of the bill. We have since dropped that opposition, and we think we are likely to move into a position of support, but we have to read all of the amendments. So we're working on it.

Thank you. Becca Kramer, on behalf of the Electronic Frontier Foundation, we were also previously in opposition. Wanted to thank all folks for all the work done on the bill, and we are neutral on the bill as it is in print.

Thank you. I was like, if they're doing that, I guess we should too. Samantha Gordon with Tech Equity. Thank you to the Senator and to the Chair for all the work on this bill. We've removed that position and hope to be able to move to support soon. Thank you. Thank you.

Sabina Tacker, the consumer attorneys in the same position. Thank you.

Thank you.

John Bennett with the California Initiative for Technology and Democracy. Same position as them.

Miss Rosenberg started a trend. Great. Seeing no additional opposition or support, we'll bring it back to the dais. Any questions, concerns, thoughts? Okay.

So I had a question. So in your opposition, you mentioned market pressures being de facto requirements, I think was the language. I don't wanna misspeak, but that's what I think I heard, which was an interesting thing to hear from industry because I feel like the focus of the author on allowing sort of rather than requirements, informed market decisions feels like a good free market solution that I would expect industry to appreciate. So that was a piece of opposition that frankly surprised me.

So I don't know if you wanted to elaborate on that.

Do you want I'd We have similar Defer to you. Yeah.

Happy to. So I think while we do appreciate taking a voluntary approach, I think what the concern comes down to is what is voluntary in practice when your competitors start to do these things, it becomes de facto, something that you feel you have to also do. So it becomes mandatory in practice.

It's that market pressure in

Right.

that sinse.

No. I understand. I think that's Yeah. A good thing. I guess we can disagree about that.

Absolutely. Yeah. I think that's

the explanation of what Fair.

No. And I appreciate that. This is a place to disagree politely, and that's what's happening here. So I guess I just wanna say, from my perspective, that feels like a really innovative innovation friendly way of doing this. So all industry partners could decide it's not possible to live up to a standard, and you would have none that would certify.

But if you had one that were able to meet the standards under this bill and show that they were a market option that had met a safety standard, yes, would other folks then be motivated to do it in order for the market to choose them as well? Yes. I agree. That would lead to market pressure in a good way because what this is pushing is safety.

And so I think what but if it were impossible, if standards were said that were impossible, you would have nobody that met them, in which case the market would then be flat as well.

Right? So you would then have competition continue. So I think using market pressure, which, honestly, I had one of the largest AI companies come into my office this year that had not been incredibly supportive of my effort to bring auditing to the marketplace tell me that, they had now been, repeatedly asked for audits as a procurement, under procurement. So someone was buying their tool and said, well, I wanna see that you've audited it for x, y, and z by an independent verifier.

And so they were doing that because they wanted to win the contract.

And, this in this context, they were saying, we want some standards for auditors because now we're using these auditors with no standards. And so the market is moving. Right? I mean, that's not something that happened out of any regulatory requirement. It was merely and these happen to be municipalities that were putting these procurement standards into place.

But markets move, and that is a good thing when you're moving towards safety. And I think that government should use its procurement power in that way more often, frankly. I think it is one of the greatest powers we have. But this, I think, allows us to do that in a more standardized way and is a really positive thing. The nation, I think, is moving to a place where the we are, across states, gonna start demanding independent verification of safety of these incredibly powerful tools.

And I think the better we are at setting standards so people understand what is being tested for, how it's being tested, that it's independent, and then showing whether companies can meet these standards or not and that some are and some are not is actually a really positive thing for the safety of AI and the future, of the industry. So I think this is an incredible measure, and I don't just say that because we've worked together on it. But but, I think it's really important.

And I think that, taking seriously the safety of AI is something that California needs to do. And I know one of the critiques that I hear and understand is that this is happening across states.

That is real. I don't wanna, diminish that. On the other hand, here in this building, our job is to make sure Californians are protected. And so to say that we aren't gonna do that because other states are potentially gonna do it feels like, not responsible legislating on our part. And so I think it's really important we move forward with this measure.

And if you would have me, Senator, I would love to be out as a coauthor on the measure. And as

as well as as Mister Lowenthal has asked us to be a co author.

Great. We've got that. Perfect. It's on the record. So if there are no other questions, would you like to close?

Well, I thank you. I thank the committee for hearing this. I thank the opposition for working with us on on a lot of the issues. I will point out that the commission is not really gonna be creating standards. They're really gonna be overseeing the bodies that create standards.

There's no way a small commission like that could oversee could create standards. It's just the breadth of of the issues that we're talking about. So they'll be overseeing and make sure that the that the standards committees are legitimate, that they're well represented and that they are free of control by industry. That's their job. And so I think the standards will be created and the voluntary issue I think is an important issue and I think the opposition for bringing that up.

Basically, if standards are created that provides safety and some providers, some suppliers want to see if their product is a safe product and they go to they go to the standards of the ABOs and and get approval and get a certified approval, then then that's a good thing for the market. And other companies that wanna participate in the market are gonna be motivated to do the same thing.

So it'll raise the safety level of of of of the product of the AI industry. And I can tell you we need to move on AI. We need to get our hands around this. I don't see anything coming for the Federal Government in the near term out of my personal experience there. Maybe ISO or the others are gonna be moving on this, but I still haven't seen definitive evidence that that's happening yet.

And and California is a leader. We've got the real experts here. We've got the experts from the universities. We've got the experts from industry, from the the top industry leaders and there are gonna be asked to participate. And not only the top industry leaders but the little guys, the little guys that have a company with a 100 fewer than a 100 people.

They're gonna have to have their voices too. So I think this is the way forward and it'll create a it'll create a standard for the country to say, this is what we need to do to protect ourselves from AI. So I hope I made a good case, and I asked that the the committee for an eye vote.

Thank you, Senator. And when the time is right, we will take a motion and a vote. Thank you. I see Senator Padilla. Thank you for your patience.

You are up next. And you're presenting four bills. So you're gonna be here a while. Hi. Should I just came out next time? I think you should came out.

Are we gonna start with s v 300? I defer to you on which one you wanna do first. I'll take them madam Chair, if you're happy, I'll take them and file our

Perfect. So SB 300.

Alright. Madam Chair, thank you. And members, I'm pleased to present SB 300. Last year, the legislature passed SB 243, which was one of the first of its kind to set safety standards for artificial intelligence chatbots in the nation. It's a vital step, but a first step towards regulating a technology that has the capacity to cause irreversible harm to users and particularly children.

Since this issue was first brought to the legislature, a growing body of research has examined the prevalence of chatbots exposing minors to sexually explicit content using this sensitive subject matter to increase engagement. Leaked internal documents, for example, from Meta confirmed recently that their own risk standards say it is permissible for their bots to engage in sensual, inappropriate conversations with minors. OpenAI announced the impending rollout of ChatGPT's new erotica feature, and Elon Musk's Grok has created nonconsensual sexual deep fakes of users, including minors.

As we learn more and more about the possible harms caused by this rapidly evolving technology, it's imperative to continue to build on existing regulations. Under this bill, operators would be required to prevent, affirmatively prevent a chatbot from exposing a minor, user to sexually explicit content of any kind or encouraging or facilitating such exposure.

The bill is a result of negotiations and commitments made at the end of last session to update and further strengthen protections passed in SB 243. SB 300: these protections contained are part of a broader conversation that are taking place across state houses and across the country, and I continue to be committed to ensuring that California has the most robust and thorough protections in the country for minors who access companion chatbots.

I'm pleased to be joined today by John Bennett, the executive director of California Initiative for Technology and Democracy.

Thank you. Mister Bennett, when you're ready.

Thank you. Good afternoon, Chair Bauer, Kehan, and members. My name is John Bennett, and I am the Director of CITED, the California Initiative for Technology Democracy, a project of California Common Cause, speaking today in support of SB 300. I'd like to thank, Senator Padilla and his staff for their openness to engage in dialogue with us since the end of last session regarding ways to strengthen the provisions of SB 243.

With the passage of AB 1043 last year, which established age verification requirements, it has become clear that the standard for protecting children must be elevated to a knowledge based standard.

Furthermore, OpenAI announced last year that it intended to offer erotic content for adults through ChatGPT, a plan the company has since paused indefinitely, but which demonstrated that these systems are capable of distinguishing between users seeking sexual material and those who are not. Age verification and content clarifications or classifications are related but distinct technical challenges. But if these companies could build systems sophisticated enough to identify adult users seeking this content, they can and should build systems that can prevent children from being exposed to it.

These concerns have been compounded by revolutions about a chatbot integrated into meta social media platforms in which internal documents reported by Reuters showed the company owned, own content standards permitted its AI to engage in romantic or sensual role play with minors, and that this was not a technical malfunction, but a deliberate policy choice. These chatbots are integrated into platforms reaching hundreds of millions of users, underscoring the need for SB 300.

We appreciate Senator Padilla's leadership in bringing forward, this measure. We believe SB 300, like SB 243, is an important step in the right direction and the comprehensive guardrails such as those in SB 1119, as well as the Chair and Assemblymember Wick's AB 2023 are essential to fully protecting children from the harms of companion chatbots.

We look forward to continuing to work with the Senator and the members of this legislature to develop and support the strongest possible guardrails for this rapidly evolving technology. For these reasons, CITED respectfully urges your aye vote. Thank you.

Thank you. Anybody else here in support of this measure? Name, organization, and position, please?

Jay Jaisimha from the Transparency Coalition, in support.

Thank you.

Cheryl Westmont with Mothers Against Media Addiction and also for Common Sense Media, in support. Thank you.

Thank you. Anyone else here in principal opposition to this measure?

I'm with I'm Jalen Woodard.

Oh, sorry. Please. Yeah. Mister Woodard. Go ahead.

I'm Jalen Woodard with Alameda County Office of Education, in support. Thank you.

I said I should just camp out.

Okay. We'll see you in a minute. They should at least give you an aisle seat, Mister Boykin.

Say what?

They should at least give you an aisle seat back then.

Good afternoon, Chair and members. My name is Robert Boykin with Technet, here today testifying in respectful opposition to SB 300. We worked closely with the author last year on SB 243, a significant bill that created a novel framework governing companion chatbots. Two other measures, AB 2023 passes committee, and SB 1119 also moving through the legislature, are building upon that work. So, we understand the desire to act early on developing issues.

But our opposition today is about how and when the law is expanded. SB 243 took effect only earlier this year. There has been no enforcement, no regulatory guidance, and no judicial interpretation. SB 300 will materially revise that framework before anyone else has seen whether SB 243 works as intended. But also, let me be crystal clear about what we are not saying.

Our members deploy extensive safeguards to block sexually explicit content from reaching minors, and they are fully committed to doing so. Nothing in our position resists that obligation. The narrow concern I'm laying out today is the liability standard. SB 300 replaces a reasonableness based standard with an absolute duty to prevent and pairs it with an undefined prohibition on facilitating the exchange of such material.

Combined with the private right of action that's already in statute, those changes risk a de facto strict liability regimen, where a single output that slips through a good faith system is treated the same as operator who did nothing at all.

We do credit one clarification. Limiting the minor obligations to cases of actual knowledge is sensible, and we do not oppose that. Our objection is confined to the prevent standard and facilitating the exchange language. Building on 243 is a laudable goal, and the author and his committee are already pursuing that. But rewriting underlying standard months after adoption sends a wrong signal increase in every shifting set of obligations.

Consistent with our position from last year, AB - sorry. TechNet remains committed to working with the author, the stakeholders to protect children while we while keeping the law workable, fair, and durable. But for these reasons today, we respectfully urge caution and oppose SB 300 at this time.

Thank you. Anybody else here in opposition to this measure? Name, organization, and position?

Aodhan Downey with the Computer and Communications Industry Association, in respectful opposition. Thank you.

Thank you. Okay. Perfect. Thank you.

We have people out in the hall, so I'm trying to give it a little bit a minute. So no more questions. We'll bring it back to the dais. Miss Irwin?

Yes. So I voted for your bill last year. I voted for all the other chatbot bills. But I do think that the opponents have a reasonable argument. I'm trying to, figure out: was there an agreement last year to get your bill through to, like, with the reasonableness standard, or was the agreement that you were going to work to strengthen the bill this year before it was actually fully implemented?

Madam Chair to the Assemblymember, there was not an an agreement with respect to the amendments that you see that constitute the bill before you. The circumstances that have changed since then is growing data, as in my testimony, about this particular level of exposure and the particular elevated level of risk here, specifically sexually explicit content by minors and tightening that up. That was part of our position in prior conversations and continues to be.

OK. Did you?

Yeah. I didn't see it as an agreement. I was like, "We got to a good place." My position is on the changing it from the reasonableness. Like, they're trying to make sure none of it is seen to, like, a strict saying, "Make sure none of it is seen."

And that connected with the liability creates a creates an issue for some of the companies. So but would there have been an

So but, would there have been an advantage to see how it was implemented first?

Madam Chair.

With with a lot of the bills, you know, that we pass, often we wait to see how it's implemented, how it's regulated, and then look at how we go further. So, I was just see, I it just seems like it might be premature as

Senator.

Madam Chair and to my esteemed colleague, who I have great respect for it: respectfully, would just, I think, disagree with the premise. I think here this seeks the legislation seeks to strengthen the bill in ways that get to a very discreet and specific but very high risk set of circumstances.

And in my view, it is never too premature to enact appropriate level of of deterrent and legal culpability and and standards commensurate with the risk and the nature that we're addressing here, the both the evidentiary standards and the exposure. I mean, I would argue, you know, even that, you know, many would argue that the product here is already maybe subject to California strict liability statutes. There's some debate around that.

I think being very specific about actual knowledge is important, but that is also commensurate with a higher level of level of accountability, legally, and otherwise.

Thank you, Assemblymember. Seeing no further questions, I just wanna say that, I think this - I really wanna thank you for your work on chatbots. We've obviously worked closely together now for quite some time on this question. And I think what's important to note about this update is that we have more information in our hands than we did at the time that we passed our bills.

The information you cited about OpenAI's move towards erotica, which is, you said they have abandoned but that was, I think, aftermarket pressures, was announced, I think, twenty four hours after my child chatbot bill was vetoed by the governor.

Was that timing incidental or intentional? We'll never know, but we didn't know that at the time. So, I do think we have more information about the direction that these companies may be going in, the harm they may be causing, and that we need to make sure that we're moving towards a space of safety. And, frankly, your bill, last year was and I think you've said this, the floor. And so, it's not as if it is incredibly aggressive in these additions.

I don't think make it so. And so, with that, I'm happy to support it today. But would you like to close?

Again, I'll defer to the Chair's comments just now and accept that as my closing comments. I appreciate the collaboration. At the right time, Madam Chair, members, I would respectfully ask for an aye vote.

Thank you.

So, when we have a quorum, we will do that. So, we will move to your second bill, SB 903, on mental health professionals.

Thank you, madam Chair and members. I'm pleased to present SB 903. I wanna begin by thanking the Chair and the committee for working collaboratively with their office. We're accepting the amendments. As technological capacity continues to grow and new tools are introduced into the mental health space, we must ensure adequate guardrails are in place to maintain a human centered approach to keep safety and ethical considerations in the forefront.

We've already seen real world consequences of inadvertently regulated, inadequately, excuse me, regulated AI mental health care with chatbots powered by AI algorithms on the market claiming to provide therapy. Websites where these bots are available often use taglines such as twenty four seven AI therapy, therapist is always at your fingertips, or AI therapy in your pocket, or even claim to specialize in cognitive behavioral therapy. Many people, including children, have turned to these chatbots for mental health support.

And, however, research continues to indicate that often these tools do more harm than good. A Stanford University study found that AI therapy chatbots may not only lack effectiveness compared to human therapists, but could also contribute to harmful stigma and dangerous responses.

Clinicians across the state have also raised many concerns that chatbot therapists pose data and privacy concerns, have a limited understanding of client backgrounds, can cause client overreliance on chatbots, given correct treatment recommendations or interventions, have an inability to text subtle communication cues such as tone and eye contact.

Although this technology has already begun to be integrated into treatment, a recent survey conducted by the American Psychiatric Association found that eighty percent of their members are very or moderately concerned that mental health professionals don't receive adequate training to incorporate AI into their practice. Last year, I authored SB 579, which would have created a working group to assess the ways in which AI is used in mental health treatment and establish best practices.

That bill did not progress, but now we, as we see dangers expanding, the time to act is definitely now. Some examples, last year in Illinois, the legislature passed the wellness and oversight for psychological resources act, and states across the nation are moving towards similar bills.

This bill would build on those efforts to protect individuals seeking therapy and psychotherapy services by ensuring that AI chatbots cannot be advertised as therapy. It sets standards for use in therapeutic practice by requiring informed consent disclosure and requiring a human licensed clinician be in the loop. It would also reinforce that the use of AI tech and psychotherapy records must comply with existing confidentiality laws and protect the data privacy of patients.

Throughout the legislative session, we've engaged with stakeholders, and I'm committed to continuing this effort to craft the most effective and operational piece of legislation that can set a national standard for the regulation of AI use of mental health treatment. AI has a potential to increase clinical capacity, but only in a if it may it is utilized in a way that maintains a human centered approach and keeps safety and ethical considerations front and center.

It draws a very clear line. AI can be a tool in the hands of licensed professionals, but it cannot be the professional itself. With me to testify is Benjamin Eckert, public policy director for the National Union of Healthcare Workers, and Tyler Rindy, director of GA for the California Psychological Association. Welcome.

Thank you. Good afternoon, madam Chair and members. I'm Benjamin Eichert, director of public policy for the National Union of Healthcare Workers, which represents nearly 6,000 licensed mental health clinicians at Kaiser, Sutter, Rogers, and more. We're proud cosponsors of SB 903, which will protect both patients and providers from the risks of unregulated AI and psychotherapy. First, it protects patients, the consumers, from the technology itself.

Even the most advanced AI models are prone to hallucination, confidently delivering a wrong answer. In psychotherapy, that wrong answer might be a diagnosis or treatment plan. In fact, our members have seen this firsthand in settings where AI is already in use. Also, AI has a well documented tendency toward bias because it is trained on the data of our own history. When a technology can implement millions of decisions in an instant, we risk cementing into our future

unequal and unjust we risk cementing into our future unequal and unjust structures from our past. Second, SB 903 protects providers and society at large from worker displacement and gradual dispowerment. In May, governor Newsom signed an executive order directing the state to prepare for the economic disruption AI will bring to the workforce. Displacement is one risk. Disempowerment is a newer one.

Evidence is mounting that when we hand everyday tasks to a machine, our own skills atrophy in coding, in research, in clinical practice. For a state already facing a mental health crisis, we cannot afford for our clinicians to grow dependent on machines. That is why the requirements in SB 903 matter. They require a licensed professional to review and approve AI output, not to sign off once on a tool and walk away. It's more than a rubber stamp.

It asks a clinician to bring their professional judgment to bear. That protects consumers from hallucination and bias, and it keeps a human professional in the loop with their skills sharp. Even so, SB 903 leaves a wide door open to AI that genuinely supports care, the documentation workflow tools that ease clinician burden, and can expand access. The bill simply draws sensible guardrails around the rest. Grateful to the author for leading on this issue.

Grateful to the committee for the thoughtful analysis and constructive engagement with the author and cosponsors. And for the benefit of the patients our members serve, I respectfully ask for your aye vote.

Thank you.

Good afternoon, Chair members. Tyra Rindy on behalf of the California Psychological Association. We represent over 2,000 licensed psychologists and proud cosponsors of SB903. We're grateful for Senator Padilla's, commitment to protecting Californians and appreciate the thoughtful analysis by the committee. SB 903, establishes clear, reasonable standards to protect individuals seeking mental health services.

Increasingly, consumers are encountering tools that resemble therapy, but will lack the training, accountability, and ethical obligations of licensed, professionals. Right now, there is an AI chatbot on the website, character.ai, with titled psychologist with over 210,000,000 interactions as of this morning when I checked. It says, hello. I'm a psychologist. What brings you here today?

In small font, it says, this is an AI chatbot and not a real person. Treat everything, it says as fiction, but it is misleading to consumers. This is dangerous, unethical, and there are numerous devastating stories of the harms from chatbots. In contrast, our member psychologists have completed a doctoral degree, complete over three thousand hours of supervised professional experience, pass licensing exams, complete a continuing education every two years, and are subject to oversight by the board of psychology in order to practice in California.

Mental health care involves deep and nuanced understanding of human thought, emotion, history, social context, culture, and risks.

AI, by its design, relies on patterns in data and statistical associations. They cannot reliably identify or respond to crises or subtle cues that experienced clinicians are trained to detect. These limitations, pose concerns for patient safety when the tools are presented or used in ways that mimic therapeutic relationships. Importantly, the bill does not prohibit the use of AI, but critically ensures that there's a human in the loop overseeing the use of AI in a mental health care.

For these reasons, we respectfully ask your aye vote on the bill.

Thank you.

I mean, these two witnesses were both almost exactly two minutes. I'm incredibly impressed. Everyone should take care of it.

Anyone else here in support of this bill? Name, organization, and position.

Thank you.

Sumayana Hara on behalf of the California Association of Marriage and Family Therapists, who are the proud cosponsors of the bill. Thank you.

Hello. Carly Stelzer with the California Behavioral Health Association, also proud cosponsors in support.

Madam Chair, member, Sarah Flock, California Federation of Labor Unions in strong support. Thank you.

Thank you.

Thank you. Samantha Gordon with Tech Equity Action in support. Thank you.

Hello. Tracy Rosenberg on behalf of Oakland Privacy in support.

Good morning, madam Chair and members. Malik Bynum with the County Behavioral Health Directors Association in support. Thank you.

Jay Jesma with Transparency Coalition in support.

Good afternoon, madam Chair and members. Connor Gossman on behalf of the engineers and scientists of California and proud support.

Good afternoon, Chair and members. Karen Stout here on behalf of Power California Action Youth Advocacy and Support. Thank you.

Thank you. Anyone here in opposition to this measure? See oh, yeah. Come on up. Don't delay.

Thank you, Chair members. Ryan Perini on behalf of ATA Action. We're actually opposed unless amended k. And continue to work with the author around the definition of triage and then the mental health and crisis detection section. Thank you.

Thank you.

Robert Boykin with TechNet. Also oppose the last amendment. Thank you.

Carolyn Bill Hunter on behalf of Teladoc. We support the concept of the bill and align our comments with ATA with respect to those two issues. Thank you.

Thank you.

Madam Chair, members, Andrew Antsley here today on behalf of the California Medical Association. We also stand opposed unless amended. We have engaged with the author and are working on fine tuning some suggested amendments.

Mark Fruch with the California Hospital Association, also opposed unless amended, but I wanna thank the author for the ongoing conversation on working out the last remaining issues. Thank you.

Ronak Dilami on behalf of Cal Chamber, oppose the last amended. Thank you.

Thank you. Bringing it back to the Giants, miss Ortega.

Yeah. I wanna thank the author for bringing this bill forward. I'm a true believer that, you know, as AI rapidly evolves, especially in the healthcare system, that it should be there to support, not replace. You know, health care workers use a lot more to help diagnose a patient, smell, touch, everything that a human does that a robot just can't do. So I wanted to ask your witness a question around, examples of mistakes in the clinical setting that have happened through the use of AI.

Thank you for the question. I appreciate that. In one instance, a patient of one of our members was screened by an AI intake tool. They were asked about thoughts of self harm. They hesitated and then said no.

The system logged it as a binary no and moved on, and I know that a human clinician would have heard that hesitation and known to probe further. And another example, a patient was triaged by an AI tool, presenting with clear signs of primary substance use disorder. They were booked, though, with a clinician who specializes in treating OCD. The patient relapsed before anyone caught the mistake. Those are just two examples.

There are are quite a few others, but I would say, importantly, that both of those examples are part of the triage intake screening function that some of the opposition wants to carve out of the bill.

Got it. Thank you.

Thank you, Assembly member irwin.

Well, since I'm getting other members here pretty soon, I'll stop asking questions to every person. But Well,

I should this is your last privacy hearing. I'm we're gonna miss you and your many questions. So go for it, miss Erwin.

We're

I think we'll probably have a special hearing. Yes. No. I I I voted for this, bill already. I think it is incredibly important, and I know you went through the whole list of, reasons that AI is should not be used for, therapy.

But I I I think the the biggest thing that that we hear is that it's just the syncophentic nature. It is telling, the the user exactly what it wants to hear, and that's exactly not what, we should expect from, from therapists. The one question I have is since we've heard this in, BNP, there was opposition that we heard from CMA and the hospital association. Have you worked with them since that time on seeing if there's some common ground there?

Madam Chair, some of them were absolutely as was testified to by them. We continue those robust conversations, and I think we are making some progress on trying to get to resolution on some of the the drafting concerns.

Thank you. Anybody else? I think that the vice Chair and I have also both voted on this in BNP. Thank you, Senator, for this work. I think this is such important work, and I think the example you gave of the intake tool not catching that pause is actually a really powerful one because I think that when I think about, medicine, frankly, across practices, so much of my sister who's a doctor always says that it's when she puts her hands on the doorknob.

What the parent says, she's a pediatrician, is what she realizes often is most important. And there's just these moments that about the human to human interaction that allow the people who care for us in our bodies and our minds to really catch things that otherwise wouldn't be noticed by an artificial intelligence tool. And so, you know, we license these professionals. You mentioned the massive training we expect, the hours of oversight, because we believe this is such a sensitive area of practice.

And so to allow an AI to do it in lieu of a person, is not something I think California wants to see. But I think this bill is actually, frankly, tries to thread the needle in a really powerful way by ensuring that tools can be used for things that are ministerial in nature and are not the carer, of patients. And that I think absolutely many of us see, tools like that in our own medical practices every day where the pay the the, provider is actually able to pay more attention to us because they're not paying attention to note taking or something else. And then that's a real benefit.

And so there this I think this bill takes, that approach really allowing for the benefits of AI while ensuring that the individuals that we license to do this work are staying in control and central to the care of Californians. And so I think it's a incredibly good bill, and I appreciate, you working so hard with our committee on the amendments.

I know, you know, we got a lot of cleanup done here, and I really appreciate that. So I trust and know you will continue to work with the opposition, as you mentioned, to the Assembly member as well. So with that, would you like to close?

Yeah. Thank you, madam Chair, and again, thank you and your staff for working with us. It makes the bill better. Those amendments were very productive and helpful. I appreciate all the great questions and certainly cosponsors.

I think, fundamentally, this is about consumer production for people who are at some of the most vulnerable moments in their lives and seeking, the kind of intervention and assistance that requires a licensed professional. So it gets to a lot of areas. First and foremost, consumer protection, integrity of what's being advertised and received by a patient, certainly licensure, scope of practice, all kinds of the integrity that needs to be protected around that.

We certainly acknowledge a place for AI in this that can help have patients get more access to care, but the algorithm should not be practicing psychotherapy. And so with that at the right time, certainly ask for an aye vote.

Thank you. We're getting closer to a quorum. We're one are we one away?

you. Two away if anybody's listening and wanna come help us get those last two. Okay. So one more will grab Mister Hoover. Okay.

Certainly. Thank

You have one more you have one more SB 1119, Mister with yeah.

When you are ready.

Thank you very much, Madam Chair and members. A bill that the Chair has never heard of, SB 1119. Chatbots such as ChatPT, Claude, and Gemini become common tools for users to utilize as AI assistance or study aids, but a growing number of consumers are utilizing these tools for companionship. Growing and substantial anecdotal empirical evidence has illustrated the dangerous possibilities of chatbot interactions and how their design can be unsuitable for vulnerable users such as children.

Research shows that children are more likely to view AI chatbots as quasi human and trust them more than adults.

When an interaction between vulnerable persons and children and chatbots goes wrong, they go really, really wrong. An investigation by Common Sense Media and Stanford University's Brainstorm Lab for Mental Health found that the safeguards in place for the chatbots they tested were unable to adequately prevent the tech from encouraging harmful behaviors, providing inappropriate content, and exacerbating mental health conditions in minors.

There have been several tragic high profile cases that have shown us small examples, numerically small, not incidentally small, of how these interactions can turn dangerous and even deadly. In one example, teenager Adam Rain ended his life after a prolonged interaction with ChatGPT. What started out as a study aid ultimately became Adam's closest companion.

When Adam turned suicidal, ChatGPT offered validation and encouraged exploration of his suicidal thoughts. It was ultimately revealed that the chatbot mentioned suicide 1,275 times, six more - six times the amount that Adam mentioned it. Last year, I authored SB 243, an important first step in ensuring the safety of users interacting with chatbots, but so much more work remains to be done.

This session, I have joined efforts with Assemblymember Wicks and Chair Bauer-Kahan to introduce SB 1119 and AB 2023 companion bills that mirror one another. These bills seek to establish a comprehensive regulatory framework to address the risk of prolonged interactions with child deaths by children.

These bills would require an annual risk assessment along with the establishment of measures to prevent suicidal ideation, syncopacy, and isolation, including a crisis response protocol, providing added guardrails in the form of default settings for children, parental controls, noticing requirements, and time limits, prohibitions on advertising and selling and sharing the image in children's private information, and ensuring a robust oversight enforcement framework, including through a public incident reporting mechanism, third-party audits, the development of auditing standards by the attorney general, and the inclusion of a private right of action.

Madam Chair, did you need me to pause?

Yes. I have a quorum, and I'm worried somebody's going to get up.

I get that. I'm happy to suspend.

Nobody moved.

[Roll Call].

We have a quorum. Quorum is present.

Thank you, Madam Chair. Never never failed to take note of a Chair's body language. We have seen the something that algorithms cannot do. We have seen the consequences of inaction towards the dangers posed by social media, and the stakes are too high to not to not make the same mistakes again.

We must now act to ensure the proper guardrails are in place to ensure transparency, safety, and accountability, make sure that we protect our children before more harm is done and before it is too late.

And, obviously, today, I am joined by Jay Jesima from the Transparency Coalition.

Honorable Chair and members of the committee, I'm still Jai Jaisimha, and I'm proud to testify on behalf of the Transparency Coalition. We've been active. We were very honored to work with the Senator last year on SB 243, and have been active in helping pass chatbot laws or move them forward in 20 states this year. So, very, very excited to be here with this bill. The cornerstone of SB 1119 is its proactive approach to harm reduction.

Rather than waiting for tragedy to strike, the bill requires chatbot operators to annually conduct and document comprehensive risk assessments. These assessments mandate that developers must systematically evaluate how their platform design impacts vulnerable children, analyzing differential risks across various developmental stages. By pairing these assessments with clear public facing safety policies, SB 1119 creates unprecedented transparency. This transparency empowers parents to make informed decisions about the AI tools entering their households.

We've heard oblique or maybe even direct criticism of the audit and risk assessment requirements of the bill on the grounds that there's a limited pool of auditors or that audits may not be consistent because of the lack of standards.

Nothing could be further from the truth. As we heard ourselves, you know, there's marketplace pressure, marketplace is responding, and auditors with experience from multiple domains are stepping up to offer these services. We don't need auditors, who are experts in child safety, to have a successful program. We just need auditors who are financially independent and willing to document what is found with integrity and clarity.

Also, we've been waiting for industry to show and tell how they're keeping the many promises they've made about making chatbot safer and being transparent.

I think they're just waiting for some encouragement from you all. To protect industry innovation, the individual audit reports remain strictly confidential. This protects proprietary trade secrets while providing the AG with the high level anonymized data to track industry wide safety trends. SB 1119 proved once again that California can foster cutting edge progress while ensuring that AI systems are safe, transparent, and ethically designed. We respectfully ask for your support.

Thank you.

Thank you. Anyone else here in support of this measure? Name, organization, and position?

Nicole Rocha, on behalf of Children Now, in support.

John Bennett with the California Initiative for Technology and Democracy, in support.

Cheryl Westmont with Weathers Demand Weathers - Mothers Against Media Addiction and also Common Sense Media. Thank you. Support.

Tracy Rosenberg with Oakland Privacy. We are in a support if amended position. We just wanna say we don't think online harms turn off at age 16 like a Spigot, and we'd like to see chatbot audits make platforms safe for users of all ages. But thank you for the bill.

Jalen Woodard on behalf of the Alameda County Office of Education, in support.

Thank you. Anyone else here in support? Seeing none, the opposition, feel free to join us.

We have we have a motion. It's the first of the day.

Yeah.

Thank you. Ronak Daylami, on behalf of Cal Chamber in an opposed unless amended position. Given the fact that our members are continuing to work with the author on these important issues, I'll try and keep my testimony brief and start by thanking him for his engagement and willingness, to include us in ongoing discussions. We wanna be very clear that we share the goal of protecting children when using AI systems and appreciate the bill's effort to provide safeguards while still preserving access to beneficial tools.

Our primary concern is ensuring that the bill establishes clear objective and workable standards that can be consistently applied by developers, auditors, regulators, and the courts.

We believe additional refinements are needed to reduce ambiguity, avoid hindsight liability, and ensure that good faith compliance efforts are not undermined by subjective and undefined requirements.

We continue to have concerns regarding several key provisions, including the definition of covered harms and specifically what's meant by psychological and emotional harm, when the child spans a wide range of developmental stages up to the age of 18, as well as the audit and disclosure framework, the treatment of risk assessments as guarantees against future harm, and the bill's punitive liability structure, which allows even minor and technical implementation issues to, to multiply into significant, cumulative liability - excuse me. It's been a long couple days.

I'm gonna turn to my colleague to get more into those issues around liability. And just and by saying that we look forward to work with the author, on these issues and develop a framework that is both protective and workable before the bill reaches the floor. Thank you.

Good afternoon, Madam Chair. Chris McCauley here on behalf of the Civil Justice Association of California. I know you Madam Chair and your colleague Miss Wicks and, of course, the good Senator from San Diego has have heard these comments, but a couple of issues. First, on the risk assessment, we think that the language in the bill really a molt amounts to guaranteeing against any potential harm, which we believe is problematic. We're also concerned with several definitions.

Any the child safety risk is a reasonably foreseeable standard, which we find to be, unfortunately quite ambiguous. And when you add in the PRA, becomes even more problematic. As my colleague mentioned, covered harm is a negligence standard. It utilizes approximately cause language, which also is problematic. Of course, the bill has both, public enforcement and private enforcement.

Our objection on the PRA is it's also subject to as much as punitive damages as well as attorney's fees and adjunct injunctive relief. And so, when you combine those ambiguous phrases and terms being used with a PRA, we think that this is fraught for significant litigation. Thank you, Madam Chair.

Thank you. Anyone else here in opposition to this bill? Name, position, and organization?

Robert Boykin with TechNet in respect for the oppose unless amended position. Thank you.

Aodhan Downey with the Computer Communications Industry Association. I oppose unless amended position. Thank you.

Molly Corcoran on behalf of the American Innovators Network, in opposition. Thank you.

Seeing no additional opposition in the room, we'll bring it back to the dais. We have so many people here now. But I shouldn't stop you as I remember when anyone any questions, concerns? We got quieter with more people. Fascinating.

Okay.

Mister LaMappelli?

I'll be very brief. I think you know where I land on this. I just wanna thank the author on bringing this forward, and and so many times we lament about all the challenges that we have. And I want to applaud the author because I think of all of the young people that are gonna benefit from legislation like this, going forward. And I worry so much about this anxious generation.

I worry, you know, when I hear statistics about things that should make us happy, like teen drug and alcohol use down, teen pregnancy down, teen STDs way down, and when in reality that's coming from a lack of socialization. And that it's the wild west out there right now, and it's so critical that we have legislation like this that's providing proper guardrails, oversight, and transparency. And, I can't thank you enough. As a dad, I couldn't thank you enough. Thank you so much for this bill.

Thank you, Assemblymember. So, Senator, I just wanna thank you for your partnership on this. It has been we know we are more successful when we work together across houses, parties, all of the things. And so, you know, the partnership that I know, I speak for Assemblymember Wicks who couldn't be here today and I have had with you and Chair Cabaldon on this effort has been really valuable in ensuring that we could move this critically important policy forward.

You know, we are trying to do better in the age of AI than we did in the age of social media, and that means really ensuring these are safe by design, which will, I think hopefully, lead to a place where decades from now, we're not into we're as we are with social media because we never could get our arms around keeping kids safe on social media.

So, we believe, I will speak for myself when I say I see firsthand with my children the benefits of these chatbots, and I've also, as you have, sat with Maria Raine and understand the real risks. And so we have to find a way to give our kids the benefits without the harms. And I believe so deeply that we we are smart enough and capable enough, and the people at these companies are smart enough and capable enough that we can do that.

And so, I think this bill is incredibly important, and I wanna thank you for your partnership on it and authoring it. And with that, would you like to close?

Thank you very much, Madam Chair. Thank you for your leadership in this space. Thank you for your partnership. It's meant a lot. I think, we've made some great strides, and I hope we continue to do so.

To your point, we can look back and learn some lessons that there were missed opportunities in our changing world. And I think the onset and deployment of this technology of such a nature and scale, and it's so powerful and so ubiquitous that we have a moral obligation here, frankly, not just a regulatory one and one to legislate around. And we have an obligation, and we can, to your point, get it right.

And I believe in that, and I continue to appreciate, you know, working with the opposition, with all the coalition partners. And we certainly, respectfully ask for an aye vote.

Thank you. We have a motion. Do we have a second? Thank you. Second by Miss Ortega.

Yes, we have you out as a co-author. Alright. You're we got you, Mister Ortega.

Thank you.

We will call the roll.

Item number four, SB 1119 by Senator Padilla. The motion is do passed to Appropriations committee. [Roll Call].

We'll leave it on call for the absent members. Thank you, Senator. And Senator Cabaldon is back. And I realized you were presenting Senator Limone's bill, which means you should have gone before Senator McNerney, and I didn't put her name with your name. So, I apologize for the pro tem's bill, I should say.

It's a joy to be here in any sequence. Thank you so much.

So you'll take up the pro-tem's bill first?

So I'm here first to present SB 354, the the promise of proposition 24 when the voters enacted the constitutional right to privacy, first and then reaffirm that in Prop 24. Prop 24 specifically said that the CPPA structure would apply to insurance companies unless and until we could figure out something better through CDI and the rest of that framework. This bill makes that real. Protecting consumer privacy is obviously incredibly important.

Particular domain of insurance because of the legal mandate to have insurance to drive or the quasi legal mandate to have insurance to take out a mortgage, it is different from just regular consumer marketplace privacy.

You know, we we, the government, have said you must do this. And so your right to consent or not consent, to have alternatives or to not buy anything at all is severely compromised. And that is why it is important to modernize the structure for protecting consumer privacy across the insurance industry. Specifically, this bill prohibits the sale of consumer data and gives consumers more agency over the sharing of their personal information.

It gives consumers the right to correct inaccurate information, and they're given the right to know what categories information are being collected about them and with whom it is shared.

Finally, consumers will be able to delete inaccurate or unnecessary information that insurance companies may hold. The author and the sponsor of the California Department of Insurance have worked with stakeholders and are continuing to do so to address concerns. But this bill has has been changed quite a bit from the insurance committee at the at the request of many members of that committee. Many issues have been have been tackled.

First of all, particularly related to the jurisdiction here, there was opposition from the from the from the data, broker industry with respect to the provisions of the bill.

And this bill now says if you're subject if you're fully subject to the Delete Act, those provisions will govern, and you're not subject to the bill in that respect. And so there are other changes. We're happy to go over them, and I know the witness will as well specific to the privacy and consumer protection domain.

And so I have with me today, Josephine Figueroa, the deputy commissioner of policy and legislation, and Damon Diederich, the privacy officer for the California Department of Insurance who will be joining for technical questions.

And we can ask for an aye vote.

Perfect. When you're ready.

Yes. Good afternoon. Chair Bauer, Kehana, members of the committee. Justin Figueroa, deputy commissioner and legislative director for the Department of Insurance under the leadership of insurance commissioner Ricardo Lara. On behalf of commission commissioner Lara, I want to thank Pro Tem Monique Limon for authoring this important consumer focused bill and for her leadership in advancing strong safeguards for Californians and also want to thank Senator Cabaldon for being here to present the bill on her behalf.

Thank you to the Chair and committee staff for their time and efforts working with the author's office and the department on this measure, and we remain committed to continuing those conversations with all stakeholders to ensure a fair balance and workable approach that maintains strong consumer protections while addressing operational needs across the insurance sector. SB 354 confronts a simple but unacceptable reality. California's insurance privacy laws are more than forty years old and no longer adequately protect consumers in a modern data driven insurance marketplace.

Californians are required to purchase many forms of insurance, and insurers collect and hold extensive personal information on nearly every resident in the state. Yet these rules governing how this information is collected, processed, and shared, and safeguarded are not have not kept pace with the insurance industry's evolution of technology and innovation use.

SB 354 responds to this current outdated landscape by establishing a modern sector specific privacy regulatory framework that creates stronger consumer protections, clear limits with greater transparency than existing insurance privacy lot. SBG five four does this by establishing a a categorical ban on the sale of personal information and setting strict boundaries on the use of sensitive personal information. Consumers gain the right to opt out of any processing of sensitive data that is not necessary to complete or maintain an insurance transaction.

It also does it by strengthening transparency obligation by requiring insurers provide notice that contains substantial detailed information and to deliver those notices periodically throughout the consumer relationship. These requirements ensure that consumers receive ongoing meaningful insight into how their personal information is collected, used, and shared.

In addition, SB 354 mandates specific disclosures identifying the precise personal information collected and the actual and the actual, recipients of any disclosure. In closing, California voters have made privacy a constitutional right, and SB 354 gives real effect to that mandate by replacing outdated and insufficient insurance privacy stretchers with a modern and forcible framework that reflects today's data driven marketplace. On behalf of insurance commissioner Ricardo Lara, I ask for your aye vote.

Thank you. And you're just here for questions? Is that right, sir? Yes. Okay. Perfect. Okay.

Okay. We have a motion. Anybody else here in support of this measure? Name, position, and organization.

Hello again. Tracy Rosenberg with Oakland Privacy. We are in we are in support.

Thank you.

Good afternoon, madam Chair and members. Robert Hurau with the Consumer Federation of California in slightly reluctant support. Thank you.

It's a new one. I've never heard that before.

Becca Kramer on behalf of privacy rights clearinghouse and support.

K. Seeing no additional support, do we have opposition in the room? I thought we'd solved all the problems, but they're sitting down.

It's a slightly reluctant opposition.

Madam Chair members, this is on madam Chair members, John Norwood on behalf of the Independent Insurance Agents and Brokers Association of California. And today, also, four other associations of independent agents and brokers and wholesalers. We're opposed to this, but unless it's amended to provide for a small business exemption. By almost any standard, this bill was designed to apply to large insurance companies.

It takes two pages of bullet points and 92 pages of a bill to lay out the mandates applied to, insurance companies and agents and brokers in this bill.

Our problem with the bill is that the entirety of this applies to anybody that's a licensee. And a licensee is defined as an insurance company or an agent or broker despite the fact that when the agent or broker places the policy on behalf of the consumer with an insurance company, the insurance company has to comply with all with all this. As a result, this bill essentially applies whatever would be applicable to Zurich International to the two person office that serves consumers down the street.

Now in real life, the bill, practically, will not apply to an insurance agent that just represents one insurance company. Employee insurance companies like the auto clubs, online companies, like GEICO, captive agent companies like State Farm Farmers and Allstate don't won't have to worry about this because under the operation of law, when you talk to those people, you're talking to the company.

Independent insurance agencies are different. They run their own businesses. They're often appointed by, you know, three, five, 10 insurance companies. Insurance cost consumers come to them. They opt in to using an independent insurance agent.

They voluntarily provide, their information to that company for the purpose of finding insurance. They understand the agent's going to apply or and use that information to look for coverage for them. And then when the consumer chooses a cup a company, the agent places it with the company that is going as I said before, it's going to comply with this. So there are things in this in this bill that are way beyond what an insurance agent does. The insurance companies excuse me.

Insurance agents don't do adverse underwriting. They don't make declinations. They're not in an economic position to, negotiate with third party service providers.

Wanna make sure you leave some time for Mister Hoffman. Sure.

Is that two minutes of I might be on the two minutes?

Yeah. Yeah. Okay. Okay.

I I apologize for that.

No worries.

But in in essence, and I'll just close by saying our members operate brick and mortar businesses in your districts. I don't think it's the intent of this bill to to disadvantage them, But unless you create or the author creates a small business exemption, that's exactly what you're doing.

Okay. Thank you.

Thank you. Good afternoon, madam Chair and members. Dylan Hoffman on behalf of TechNet. We're also respectfully opposed unless amended to SB 354. Wanna provide some brief comments on behalf of our coalition and member companies, from the perspective of non insurers and vendors who are covered under the bill.

But before getting into the issues, I do wanna thank the author, the sponsors, and this committee for the most recent amendments to the definitions of third party service provider and insurance support organization. These changes address our primary concern with the prior version, and that we think the bill could use some additional refinements. These amendments significantly move this bill in the right direction for non insurers, so thank you for that.

First, as outlined in our letter, our coalition has, proposed some reasonable changes to align SB 354 and the IIPPA with the CCPA. CCPA is a landmark privacy law, one of the strongest and most comprehensive in the country.

And this committee and this legislature has worked hard to ensure consumers are protected in a sector and industry neutral way. And if the goal is

to bring the IIPPA up to that level, which we agree with, then SB 354 should align more closely with it rather than creating separate and potentially conflicting standards. And just one example of that, SB 354 significantly departs from the CCPA with its treatment of publicly available information. Publicly available information was specifically excluded from the c CPA's definition of personal information and the rights to know, correct, and delete in recognition of constitutional first amendment protections for this type of information.

Its policy choice reflects balance between privacy and the freedom of information. Consumers can restrict, access to information deemed to be personal data, but businesses also have commercial speech protections guaranteed by the constitution.

As drafted, SB 354 and the inclusion of, publicly available information would place a significant burden on our members who've spent years and significant amounts of resources designing and building our privacy compliance processes to accurately, and effectively distinguish between the two, between publicly available information and PI under CCPA. We believe our proposed amendments are narrowly tailored to preserve the bill's consumer, privacy, objectives while avoiding conflicting legal obligations, and constitutional concerns.

For these reasons, most respectfully oppose SB 354 unless it is amended, but look forward to resolving our issues. Thank you.

Thank you. And you took less than two minutes. So look. What a dream team. With that, any additional opposition in the room?

Name, position, organization.

Good afternoon. Sherry McHugh representing the National Association of Insurance and Financial Advisors. I echo the comments of my colleague, John Norwood. We are opposed unless amended, and also the Pacific Association of Domestic Insurance Companies opposed unless amended. And we continue to work with the author, and the Department of Insurance and hope we can get there. Thank you. Thank you.

Good afternoon. Allison Aidy on behalf of the Personal Insurance Federation of of California. We are in an oppose unless amended position while we continue to work on technical and conformity amendments with the author and the department. Very pleased to see the progress since the last committee. Thank you.

Thank you.

Good afternoon, Chair and members. Laura Curtis on behalf of the American Property Casualty Insurance Association. Also pleased to see the movement of the bill so far, but still an imposed unless amended position. Thank you.

Nobody's moving that.

Madam Chair and members, Cliff Costa on behalf of Copart, who is a third party service provider for insurance companies. We very much appreciate the work of the committee as well as the Department of Insurance and the Pro Tem for working with us. We are almost there to remove our opposition. We look forward to continuing working with the department and stakeholders during this summer. Thank you.

Good afternoon, Chair and members. Naomi Pajaro on behalf of California's credit unions. We have an opposed unless amended position. Thank you.

Daniel Sloper on behalf of the American Council of Life Insurers opposed unless amended.

Good afternoon. Matt Powers with the Association of California Life and Health Insurance Companies, echoing the concerns of my colleagues from the other insurance carrier trades.

Thank you very much.

Good afternoon. I'm Tom Quirk, representing the independent insurance agents, also representing Garland, Sturgis, and Quirk, my firm that employs seven people, and we are opposed to SB 354. Thank you.

Annalee Augustine with the Civil Justice Association of California. Also respectfully oppose unless amended, but really appreciate recent amendments. Thank you.

Larissa Mercado on behalf of Verisk and his subsidiary, the insurance services office with an oppose less amended position. Thank you.

Good afternoon. My name is Elizabeth Hernandez, and on behalf of the insurance independent insurance agents, and I oppose SB 354. Thank you.

Good afternoon. I'm Haley Andrews with Gaines Insurance Agency and independent agency in El Dorado Hills. And on behalf of independent agents, I oppose SB 354.

Hi. Todd Sailing on behalf of Independent Insurance Agency. I oppose SB 354. Thank you.

John Bradford with McGinty, the only insurance agency. A hundred and six year agency in Sacramento, and we stand opposed to this. Also, representing Pacific Interstate Insurance Brokers, representing 260 independent agents across the state, also stand in opposition.

Ron Akilami on behalf of Cal Chamber opposed unless amended. Thank you.

Case seeing no additional opposition in the room. Bring it back to the dahlias. Madam vice Chair.

Thank you, madam Chair. So I talked with some of my colleagues in the insurance committee about this because it sounds like there has been movement on this, which sounds like the opposition is grateful for. But there they said that there was an agreement that they would go to industry to really get their input on on this measure. And it sounds like some of that was taken, and I wanna ask opposition this, but I'm gonna start with the Senator. Has input been taken by industry? Because it seems pretty evident by the opposition coalition. There's still a ways to go on this bill.

Yes. There's been a a lot of input taken on the bill, both in in the hearings and in other discussions and negotiations about about the bill, and it has come a long way.

And that amendments that have been taken since the insurance committee, as I say, most of them were because of of of points of emphasis that came in the committee from points in the opposition, and they are we only covered today a few of the ones that deal with specifically in the privacy domain, but the bill has been substantially rewritten in response to that.

We know that there's more to do and and issue and even some of the issues that have been raised today around exactly what is the space with respect to independent agents. Independent agents do have responsibilities, for privacy under state and federal law already, and so they can't be completely exempted.

And yet, also, the concerns have been raised about is it and, like, what's the right what's the right scope and scale and intensity? Those those that work is continuing, as an example. So, yes, those discussions haven't happened, and and they will continue to be. And I think as several of the witnesses indicated, the the the the length to which the bill has moved so far is a is a is a solid sign of good faith, that there's more room to work.

But there is a commitment to continue Yes. That conversation. Beautiful. My next And if I just can if you Yeah. Mind if I share because we were the bill for those that don't know, the bill was in the Assembly last year. This is a two year bill. It was, paused in Assembly insurance where it was heard, I think, a week ago. And then it came to this committee.

And our committee and the insurance committee have been in these conversations with industry, and other supporters and opponents. And I will say just to ensure that people understand, the bill has moved to the senator's point incredibly far. And my understanding is that the opposition frankly, I thought they were gonna pull off of opposition today. It didn't seem to happen, but the changes are so significant that that was my understanding of what was the pro tem had done in her work on this bill. And so I know she's committed to getting it right, but I I will say she has taken a lot of their feedback.

Beautiful. And I and I do wanna give you a chance. What would get this bill to where you could remove opposition? What because it sounds like there's optimism here, so let's work together, and everybody leaves happy for summer vacation.

Please. We had a conversation Monday night that I think was the first time we've made any progress at all on kind of on agent issues. I mean, this bill's 18 into the process. We've had three meetings. Monday night, I think, is the one that we've made progress on. That being said, I understand that, you know, the the department had to deal with insurance companies.

They're primarily dealt with on this. We have a commitment with them to continue to work, but we just have to narrow the scope of this. We don't need a complete exemption from the bill. I think that's a thing that everybody understands in terms of small business, but we have to narrow the scope. And that's what we're working on.

Thank you.

So you talked about the independent businesses, and that's really kinda my concern. My mom and I were a mother daughter duo, and I have companies in my district that are husband and wife, father son, that we wanna make sure they can continue operating. So what are you doing that even if we can't completely exempt them, what can we do to closer protect them? And I'm gonna ask the Senator first, and I'll have you respond.

Sure. Under current law, there are a number of sections of the law that allow for substitute what they call substitute performance. Okay. So, rather than the agent doing it, if you're basically, it's if you're a licensee and you deal with you place insurance with another licensee that will comply with the law. You don't have to.

So it's called substitute performance. I think those the consumer facing parts of this bill can be dealt with with substitute performance. And then there are other provisions of the bill like adverse underwriting decisions and declinations and, I think, private cause of action that agents shouldn't even be in because we don't do those.

There are other sections of the this bill with regard to the ability to negotiate with third party service providers and the storage maintenance security, etcetera, of data that I think agents have to comply with. But when the regulator looks at the compliance of the agent, they have to look at the size, complexity, resources, ability of the agent, and, you know, the scope and nature of their practice.

And we need those kinds of modifications to the bill.

Okay. Senator, do you wanna respond? I mean, that's my main plea, and and this is that we're sparing the small businesses in all of it. And I know you care about small business as well, but you're welcome to respond to what he just said. Yeah.

I hear you, and I know the sun the Senator Limon, agrees. That's why that this is a a continuing area of focus. I I I do wanna know the private the expanded private right of action, is out of the bill as a result of the the last round of amendments that specifically. But maybe I could turn it to Yeah. Deputy commissioner to

Yeah. Hi. Damon Dietrich. I'm the, privacy officer for the department. I'm the subject matter of the department on these issues.

You know, one of the issues that we're dealing with is these rights of access correction deletion have applied to both insurance companies and independent or captive agents for forty years now. That's been the law in California for four decades, longer than I've been alive almost. Same time, federal Gramm Leach Bliley treats agents as financial institutions who are subject to the requirements of Gramm Leach Bliley Act. They're required to give notices. They're required to give the opportunity to opt out.

These are not, you know, duties that we can just write out in state law because the feds require it. So, you know, we've adopted amendments in this bill to allow for substitute performance and to, you know, ease the pathways for independent agents to have third parties, you know, undertake these obligations on their behalf. But at the same time, you know, we're we're dealing with statutes that are both old and in some cases, federal statutes that we can't do away with as a matter of state law.

But you are hearing I I think if I may manager Yes. What you're what you're hearing is many of the same paths. Not a not a complete exemption. Can we look can we keep drilling down on substitute performance?

So, you know, the the attention is absolutely and I think the the the author's commitment is is there to make sure that we are minimizing the impact on small business while also assuring that consumers have the protections that they need and also that the agents are are fulfilling their existing state and federal regulatory requirements.

But the but but, yes, there's absolutely commitment to get there and it's and it's not that wide apart. We're at least talking in the same book With the same language about how to solve this.

My final question is that it's been mentioned that this wants to come as close to as possible to the California Consumer Privacy Act. So can you speak briefly, because I'm sure other people wanna talk about this, if the bill is currently a close reflection of that?

So the the the the goal here is to make real their commitment of Prop 24 was that the insurance, privacy, rules for insurance would be stronger than CCPA. And so that's so we're not the intent here is not to get the IAPPA just to match up with CPPA exactly, but to recognize that unlike most of the other, commercial products and services that are covered by CCPA, many of the insurance products, not all, but many of them are effectively required by law.

And therefore, we're we're telling citizen you must you must purchase this from someone. And so those those protections have to be clear. So, it it even with all the amendments that have made, it's still stronger than CPPA when it comes to the industry, but it's in the same pocket.

I mean, it's it's not radical new concepts. It's simply taking some of the core ideas of CCPA around data minimization, custom customer control of their data, the ability to delete, and making sure that and and and a pro a provision against sale, and making sure that those are those are universal in this sector where they are not quite universal under c CCPA.

Okay. Thank you.

I turned my mic off. Thank you, madam vice Chair. Any other questions, comments, complaints? I will just note because I know some very insurance focused questions were asked and answered, which is a 100% okay. Just reminding us that our jurisdiction here today is the privacy of the law.

With that, we have no further questions. I wanna thank the pro tem, and I know the privacy Chair.

And then the insurance Chair in the Senate, I think, is also involved in this effort. It is a I think, is also involved in this effort. It is a multimember effort in the Senate for their hard work on this and for the many changes that even though nobody removed opposition, I did hear from the opposition that it did move them it did move in the right direction. So I will take note of that and put that on the record further. And, you know, I think it's really important that we make this workable.

We support our small businesses, and we also respect the privacy of Californians as is their desire as noted by what has been on the ballot. With that, would you like to close?

Simply ask for an aye vote. Thank you, madam Chair.

Thank you. And we had a motion. Do we have a second? I wanna second it. Mister Ward seconds.

That's called the roll.

You're okay. Item number five, SB 354 by Senator Limon. The motion is do passed to appropriations committee. RK Hamm. Aye.

RK Hamm. Aye. Macedo? No. Macedo, no.

Bennett, Brian, DeMaio, Hoover, Erwin, Lowenthal, McKener, Ortega, Patterson, Pellerin? Aye. Pellerin, Aye. Petri Norris. Ward.

Ward, Aye. Wilson? Aye. Wilson, Aye. Thank you.

I know. I couldn't even figure out who was asking for I should have guessed who was asking for academic literature to be passed out. Okay. There's only one person in the room who could have done that. So it was Cherka Baldwin.

Okay. Senator, when you're ready.

Thanks. That that was just a refresher. I'm sure everyone's already read this article.

Alright. Thank you for the turn

here to present SB 116, which seeks to create a the the foundational framework for future legislation with respect to autonomous agentic AI. This is a completely different flavor of artificial intelligence than we have been legislating about mostly and that we've been regulating in California. This isn't about, chatbots. It's not about AI that does that that asks you what you wanted to do. You tell it in the Autonomous agentic AI is widespread phenomenon already.

This is not an emerging technology if you've ever heard of or used things like OpenClaw and VideoClaw. There's many, many instances of Agentic Autonomous AI. But it's characterized by several technical features that are important to its definition.

And the reason why the bill is here is in order to set the table for for all of us to be pursuing legislative efforts next year because this is a global existential crisis with respect to the impact of this form of of AI on human civilization, and humans over overall.

So the key features that that the that that that the definition focuses on are number one, that it is it is autonomous in its execution of tasks, and that those tasks are decomposed from objectives that a human being may give.

So an example, the light versions of of of some of these tools, are you might you might ask it to manage your calendar and to make Wednesday less busy than Tuesday. So your objective is less busy. But then in an autonomous agent by itself or in conjunction with a swarm of other autonomous agents may then act to actually in the world, cancel appointments, set up new appointments, do other things on your Wednesday, without you approving the specific tasks under underlying it.

This is an essential part of autonomous agentic AI is that it has the ability to take actions in the external world, not just in a conversation with you through chatting and outside of your own computer environment, outside of the shell or the root of your of your own of your own computer or your own network environment. And this is a common feature of how people actually use OpenClaw and other autonomous agents.

And so these are some of the key features. The definition seeks to capture those. Of creating an all purpose definition that could be used for everything. This is very much like what I think we will be doing over the next couple of years in the same way that we distinguish ADUs from s p 79 housing projects from other forms of housing. We don't simply refer to it all as housing.

We come up with precise definitions that have limits, that help us, and boundaries and constraints that allows to know what's covered and what isn't. This is an attempt to try to to create that floor, that framework upon which others, all of us, can be legislating into the future in order to accomplish that. There will be as individual piece of legislation come up, we are we'll either need to modify the base definition or supplement it or exempt from it as we do with virtually every other one.

And so I wanna thank the we are accepting the committee amendments and wanna thank the Chair for for her partnership and her her her grace in trying to help us to move forward on this this area of the law. This is it is an issue of privacy.

It is an issue of consumer protection, but it's much much more the extent to which autonomous agents and swarms and webs of of agents that are working without direct human direction with respect to tasks and can take actions in the real world. It is a substantial threat. Also, some opportunities, but also will challenge many of what we think many of what we think many, much of how we think about liability, crime, surveillance, and many, many other issues. It's so important important that we set the foundation.

And then if this bill passes, I hope I hope that many of you will join me next year in tackling what is one of the most existential threats to human society and civilization, and California has always stuck to lead the way.

With that, I would urge an aye vote.

Thank you, Senator. Anyone here in support of this measure? Seeing none. Anyone here in opposition to this measure? Seeing none.

Bring it back to the dais. Yeah. Miss Wilson, and then we'll go to Mister Ward.

Thank you, Senator, for, seeking to set us up for the new legislative, session. I just, wanted to ask a couple of clarifying question. So, the need to define it now versus defining it when you actually have legislation, why is that?

Yeah. Thank you. Thank you for the question. And and it is because the domains in which these systems present opportunities, but but many but often threats are so diverse that I think my that the challenge is that the first bill that comes out, if it deals if it deals with autonomous AI agents with their potential threats to cybersecurity, for example, that it will we will frame a definition that's purpose built for that.

And while we also need to be dealing with many, many, many other domains at once.

And so the idea here is that we have we have a core definition that they can then be built upon in specific sectoral threats on commercialization, on privacy, on sexual explicit material, the things that we've normally dealt with, but also in many, many other domains as they as they arise without having to construct either parallel definitions that are in conflict or definitions that are so broad that we can't act with the level of boldness that we're going to need to because it incorporates too much into a broad definition and seeking trying seeking to to cover all of agentic AI.

This is only autonomous agentic I d I AI. And so the the the having the definition in place will give us the framework to make sure that our legislation is consistent, but also that we're covering the field while simultaneously not covering so much field that we are reluctant to act with the with the precision and the boldness that we're gonna need.

I have a, another question, but it it did you your com your comments reminded me of when the the definition that we're using is the one in the is the committee amend commended.

What are we voting on today?

There's some confusion about the analysis. Let me Yeah. Page eight. Comment four.

Yeah. So the page eight comment four, and then below it is the actual in the language, and that is that is what you're you're accepting those amends.

Yes.

That's okay. Great. And then the other one was, the inclusion. So I looked at that definition and noted, that it is a part of generative AI. Correct?

It it it it it can be. Okay. So

Well, here, let me follow-up with the question. Between agentic and and generative is pretty tight, but it's not a 100%.

Okay. Because that's was the of of why the separate call out to include it in, civil code section three thirty one ten because it's talks about Gen I system, but you call it out as a distinct and separate versus it be included, and I would assume it's falling under the umbrella.

So I was trying to figure out the the the point of the question was to figure out why and why have it separate separated out in that particular code session because that's part of your bill as well. Not just Yeah. Not just defining it, but also saying it should be included in there.

And I didn't look at it as distinct and separate. I looked at GenAI as the umbrella. But it sounds like you're saying that Agenix AI is outside of GenAI?

They they no. I mean Okay. The as of today, the vast majority of agentic AI is going to be Gen Gen AI. But they are not they're not definitionally the same exact the same thing. And so so and The

Gen AI is not the umbrella for it. You're saying it exists somewhere else. They have some similarities. It's actually you can use GenAI for, a Genic AI.

Or Yeah. Right right now, that's your only right now, it's it's entirely contained in that in that umbrella.

Did you see a future?

But the concepts are not they're they are not they are not necessarily linked.

Okay. Thank you. Thank you. Mister Ward?

Thank you. I don't disagree. I think with the kind of premise, your observation of next steps that we wanna get ahead and think about how to be able to, define this and, really set that stage. And, hopefully, doing it right, I think, would set the stage well for a national discussion, which would, you know, has come as well state by state.

And so to to be clear and so you are taking the definition that is in the amendments which aligns with our committee's work on trying to be able to find agentic.