Regulate AI

The race to regulate AI: How California could go beyond Biden

“I see the benefit of a single system of regulation, but none of us trust that D.C. can get there fast enough. AI is moving at such a rapid clip,” Bay Area Assemblymember Rebecca Bauer-Kahan told POLITICO.

By LARA KORTE

California Assemblymember Rebecca Bauer-Kahan wants the Golden State to take the lead on regulating artificial intelligence before it’s too late.

President Joe Biden this week signed an executive order marking the most ambitious national effort yet to guide a rapidly-evolving technology. The sweeping measure was praised by both industry and consumer advocates as an important first step, but without more forceful action from Congress, its impacts are likely to be limited.

Enter California — home of the globe’s most powerful tech companies and the Democrats who have become increasingly willing to go up against them.

Bauer-Kahan, a Democrat whose Bay Area district includes major tech hubs, is championing an effort to prohibit “algorithmic discrimination” — regulating automated decision tools that make a determination that may have a significant effect on a person’s life, such as in hiring, medical decisions, or parole rulings.

Her bill didn’t make it out of the Legislature this year, but it attracted the attention of tech companies and industry groups across the nation. We caught up with the lawmaker shortly after her trip to D.C. for Biden’s announcement to hear about what’s next for her legislation, and the future of AI in California.

This conversation has been edited for length and clarity

You just got back from Washington, where you met with White House leaders about AI regulation. After that trip, and Biden’s sweeping executive order, what do you see as California’s role in this arena?

“I think [Biden’s order] is very complementary … but I think our work will go a step further in ensuring that we’re creating a market for safe AI.”

Do you think Washington is looking to California to lead on this issue?

“I wouldn’t say Washington is looking to California. I think they would like to be the ones acting — there’s no question in my mind — but I don’t think they can.

I think they are open to the states acting. It does create a patchwork across the country that is really hard to comply with for these companies that will be operating across state borders. So I see the benefit of a single system of regulation, but none of us trust that D.C. can get there fast enough. AI is moving at such a rapid clip.“

AI is such a broad field. Why did you zero in on the question of algorithmic bias with Assembly Bill 331?

“I have been on the privacy committee my entire tenure, so five years now, and I had seen our previous chair, Assemblymember [Edwin] Chau, do his bills, and every single one failed on this question of ‘What is AI?’ and the definition of AI.

So I came into this thinking, ‘OK, that can’t be what stops this, because we have to protect society, so how do we think about it differently?’

To me, this felt like the lowest-hanging fruit, and the most important thing to do first because we all agree, hopefully, that we should not be discriminating in these consequential areas.

I think a much larger stab at AI is much harder, because we don’t really know where it’s going at this point.”

But even your bill, as narrow and low-hanging as it was, didn’t pass. What does that say about the path forward for regulation?

“It didn’t pass in the black hole of appropriations, so I think that is a different beast. But it did really get incredible support in committee, and every Democrat voted for it, including very moderate members, and Republicans who sat on those committees spoke very positively about it — they didn’t actually vote for it — but they were not pooh-poohing it in ways that I actually was anticipating, to be honest with you.

I do remain optimistic that this is something we can build consensus around.”

Can you tell us why you think it failed and what you’ve learned for the next round?

“It was a really tough budget year. And anything we do in this space, if we want it to be enacted in a meaningful way and we want meaningful enforcement, it will require us to build up expertise and it will require us to build up our agencies to do what they are not doing today.

A year ago, when we first started drafting legislation, ChatGPT had not even been dropped.

I think people are much more aware, even a year later, how prolific this is in our lives, and that we as a government are actually behind the ball.

There were people who would say to me, ‘This is premature,’ and I would say ‘Fifty percent of these decisions are being made by AI today’ — and that was a year ago. I imagine today it’s even more. So we really need to catch up to where the technology is going.”

California prides itself on being an incubator for innovation, but some of our tech leaders here have derided Biden’s order as stifling innovation. How do you respond to that?

“I was a regulatory lawyer at the beginning of my career, and I understand from that work that it is their job to innovate and create and to build successful businesses and employ Californians. And it is our job as the government to protect society. And I do truly believe that if we’re smart and thoughtful about how we do this, we can do both.”

What will happen if California doesn’t regulate AI?

“What I had hoped for when I introduced AB 331 was that we could set the stage for a national standard, that we wouldn’t have patchwork regulation, but we would have other states follow our lead.

But I think it is dangerous for us to let other states go first and set a standard that is not up to California standards — or that is not as nimble in allowing innovation. “

Do you think it will take another ballot initiative threat like with the privacy laws to pass AI legislation in California?

“I was on a panel a week ago and I said I don’t think we should allow that to happen. But I think that if we don’t act — I just saw a poll that 90 percent of Californians want us to regulate AI.

So if that is where the electorate is, they will act without us if we don’t act. I truly believe that.”

What can we expect from you next year in terms of AI regulation?

“This one [AB 331] will be back, I promise. It won’t be the exact bill, but it will be back.

We’re looking a lot at what is federally preempted and what isn’t.

I imagine there’s going to be a lot in this space.

And to your point about innovation and regulation, we do need to be careful to make sure we balance the two, because there’s a lot of benefit. I’ll tell you the other day I had to help my son with his math, and I used Chat GPT to help me learn how to do it.”

Did it work?

“It did, but it ... it was not advanced math.”