We Get AI for Work: Is California Regulating AI?

Takeaways

Eric J. Felsberg 

Principal and Artificial Intelligence Group Co-Leader

Hi, everyone. Welcome back to our new podcast dedicated to the issue of AI in the workplace. My name is Eric Felsberg and I'm joined by my partner Joe Lazzarotti. We've been seeing a lot of activity coming out of California as it relates to AI, and I know whenever we talk about anything having to do with California, we like to consult with one of our California partners. 

I'm happy to say that today we're joined by our partner Scott Jang, a principal in our San Francisco office. Thank you for joining us. Like I said, whenever the topic of anything having to do with California comes up, we like to have one of our California partners join. And this is no different. We are seeing a lot of activity coming out of California as it relates specifically to the topic of AI. 

So, Scott, before we get into some of the specific legislation and questions that Joe and, by extension, our listeners have, can you provide a little bit of history and context to where we find ourselves as it relates to AI in California? 

Scott P. Jang

Principal and Technology Industry Co-Leader

Absolutely. That is a good starting point when it comes to California because although legislation is currently before Governor Newsom this year, the legislative efforts and focus on AI started as early as last year. 

As some of our listeners may recall, there are actually some bills that were working through the California legislature back in 2023, having to do with adopting safety measures with respect to AI. Now in 2023, those bills ultimately did not make their way to Governor Newsom's desk because Governor Newsom had expressed some concern that California was developing kind of a patchwork of legislation to deal with AI and he wanted to address the topic in a more systemic, unified manner. So, the legislature ultimately did not pass those bills and put them before Governor Newsom back in 2023. 

But we're here at it again. There are now a raft of bills and legislation posed to Governor Newsom that are pending before him for signature and frankly may be signed in about a couple of days or so because Governor Newsom has until September 30th to sign those bills. 

Felsberg

It's interesting you mentioned the patchwork. A lot of the clients that we work with are confronting these issues. One of the things that they constantly confront is a very simple topic, which is what is the definition of AI, right? Each one of these laws around the country kind of define AI a little bit differently, right? So, if you're an employer who's trying to have national operations and you're trying to comply with all these different laws. It's a patchwork and it makes it challenging for employers to even understand whether they're subject to some of these laws and to determine that you really need to understand how a particular jurisdiction defines AI and then determine once you know that definition, whether your tool, your platform that you're interested in, falls under that definition. 

And the reason I mentioned that is one of the bills you mentioned, Scott, is targeted at defining AI, at least for purposes of certain California laws. And this is actually a very good move on behalf of the state to level the playing field, define it for everybody so that we all understand exactly what it is the state intends when it mentions AI and some of these other laws. So, Scott, can you talk a little bit about that effort that is underway out in California to define AI and how, at least right now, they're intending to define it? 

Jang

Absolutely. And this actually touches base on a discussion that we've had many times, Eric, and that is: AI in the last couple of years has almost become a buzzword that everyone says and has a general sense of, but when we get into the specifics, especially when it comes to legislative efforts at addressing AI, people can be talking about very different things. For example, are we talking about big data systems? Are we talking about generative AI? What exactly are we talking about? 

As you can imagine from a legislative standpoint, it's pretty important to make sure that we're all talking about the same thing before we're going to legislate the issue. That is also important because one of the concerns that has been consistently raised to Governor Newsom is this balance between setting a framework for developing AI technology in the future that makes practical sense while also leaving enough flexibility for those in this space to be acting with enough ingenuity and creativity that would maximize the utility of this technology.

With that said, California has actually taken a potential step at this definitional aspect of creating a framework for legislating AI. And that finds its form in AB 2885, which passed this California Assembly and is before Governor Newsom right now. AB 2885 does a couple of things:

1. It creates a definition with respect to artificial intelligence, which it defines as an engineered or machine-based system that varies in its level of autonomy and that can, for explicit or implicit objectives, infer from the input it receives how to generate outputs that can influence physical or virtual environments. 
   
   Now, that’s a lot of words. It sounds very technical, but it still underlines the point that we raised earlier, and that is: We need a basis for everyone to understand what exactly is meant by artificial intelligence. And this hopefully does exactly that.
   
   
2. AB 2885 also includes other aspects, such as requiring the state agencies to develop (i) an inventory of how current AI systems may already be implemented or being used by the government, as well as (ii) studies in terms of framework for how to legislate AI in the future. 
   
   Hopefully, this is a building block piece of legislation that California as well as potentially other states can look at with respect to how do we move forward with AI in the future.

Joseph J. Lazzarotti 

Principal and Privacy, Data and Cybersecurity Group and Artificial Intelligence & Automation Group Co-Leader

What was interesting for me, in looking at some of the bills that are running through the California legislature and now going before the governor, is the state's effort to increase transparency in terms of the disclosures and this definitional issue that you and Scott were talking about.

AB-2013, for example, provides a laundry list of things that have to be disclosed for a developer that's developing a generative artificial intelligence system of service. And what is a developer? Is it just a company that's selling a service? Is it a company that's developing something internally? Like, how does that apply? Who will it apply to under what circumstance? That may be a question for organizations.

This idea of transparency is pretty important, particularly as it concerns deepfakes and other types of uses of generative AI becoming more prolific. Oprah Winfrey had a special on ABC. If you haven't seen it, I encourage you to go look at it. It was a [number of] interesting discussions with [Sam Altman and] a lot of different people in the AI space. She was kind of amazed at some of the technology and also a little bit nervous about it. 

And it seems like not just California, but other states are approaching this issue of deep fakes and gendered AI technology a little bit differently. Tennessee has [what] they're calling the Elvis law to try to prohibit the use of someone's image without their consent for commercial purposes. So, we’re seeing some states approach it that way.

At least these laws in California seem to be approaching it, at least in part, from the standpoint of making certain disclosures about how they train their system or what steps they take to develop their tool.

SB 942, as I understand it, applies to much larger organizations that would be considered covered providers under the statute. {Having] organizations that have like a million monthly visitors to or users of their system be more transparent and even requiring them to have a tool that can help a person determine if this is some kind of a deep fake that was developed by a generative AI system. 

This whole notion of transparency is working its way through not just California's legislature, but in other states as well. It's going to be interesting to see how that all plays out. 

Jang

I agree with you, Joe. Your point about the focus on transparency is a good one to point out to our listeners, because it does circle back to that point that I raised earlier, and that is, when legislating AI and AI in the future, this delicate balance between creating a framework for the safe use of AI, but not stifling the industry with restrictions and rules that may not even be relevant two years from now. So, what you see, especially in this latest round of California legislation, is a focus on just disclosure and transparency as opposed to substantive restrictions on certain types of AI. 

And so, to your point, SB 942 goes into a requirement that certain quote unquote “covered providers” create a tool in which the public can detect AI — whenever a text, video or other multimedia content is created with generative AI — as well as a requirement that the covered provider actually make that disclosure. And I think that the Taylor Swifts and the Swifties of the world may have a better sense of when AI, generative AI, content is being used and is being floated in the public domain. And so that seems to be an area of focus for California definitely.

Lazzarotti

One other thing to get your comments on: I know some of our clients sometimes will get, either an employee providing it for some purpose, a document or photo or video, or maybe in the course of litigation in discovery, they may get a document — we're not talking about Taylor Swift and Oprah Winfrey; when the rubber hits the road for us, it may not be as glamorous — saying, “Well, is this real or not?” There's a practical issue as to how we determine that and can we rely on that at this point? I know that the resources to do that are somewhat sparse at this point.

Jang

I have to say on a couple of litigation matters, I have run into circumstances where clients have believed that text messages or photos may have been fixed or manipulated. And while there is some forensics that can be done with respect to those types of manipulated media, I anticipate with the growing use of generative AI in the public sphere that the level of sophistication of the manipulation is only going to increase. And I'm not sure whether or not the tools, the forensic tools that currently are at our disposal, necessarily are going to be sophisticated enough to go and match and detect some of the more creative versions of generative AI that might come across our desk.

Felsberg

Certainly a lot to think about, Scott. We want to thank you very much for being available to us and educating us on what is going on in California. We appreciate it and we hope that you'll keep us updated so that we can keep our listeners updated as to what's going on in California. And I'm sure we'll see a lot of similar efforts across the country thereafter. 

So, Joe, Scott, any last comments before we break?

Lazzarotti

I would only add that we're definitely looking, as we mentioned on a prior episode, if listeners have any thoughts or suggestions or want to comment on these podcasts, you can reach out to us through our email address, ai@JacksonLewis.com.

Jang

And just for me, just for listeners to put September 30 on their calendars. That's when the governor has his last day to sign these bills as well as other legislation in California. And so, we'll have a better sense of the extent that California pushes forward with its legislation on AI. 

Felsberg

Perfect. All right. Well, thank you, Scott. Thank you, Joe. And thank you all of you