What do you make of this regulation going through California? So I can be a force for good. But it does need some safety guards and some guardrails put in place. I think this is really a signal that the unregulated era of AI is over. I think that first step was with the EU Act that happened and was ratified overwhelmingly last year by the majority of the European Union. The worry for some in the market open air has been a voice for this when Wendy, is that the regulation coming through will harm in particular smaller companies and also open source large language models. Is that something that you agree with or is this the necessary step taken to ensure that AI remains for good? Well, it's a tricky regulation to put in place. There's always this balance of ensuring that we're not stifling innovation while trying to reduce harm and to create transparency in AI. Within the SB 1047 regulation itself. There are some parameters related to the

amount invested in the AI, so they're trying to find a little bit of a balance of if it is a large-scale model that is being that has, you know, over 100 million of investment or $10 million to deploy that, that is what would be subject to this as well as AI that could be used to enable cyber attacks to critical infrastructure such as energy, you know, nuclear, you know, chemicals, as an example. So it's about spending. We spoke to academics who said they were consulted but didn't like the content of the bill. We spoke to Senator Scott Wiener, who said he spoke to everyone he could and tried to get input from as many people as possible. Then you have matter that's against it. From the open-source perspective, why couldn't we find a bill that basically works for a broad coalition, given all the talk about it? It's true that it is somewhat incomplete.

I think one of the best things that can be done is that public-private partnership where, you know, companies like Meta, OpenAI, and others are really on the leading edge of developing this technology. And I think there were a few things that were basically incomplete in this. One was that there, you know, is the notion of kills, which is the other is this notion of having third parties test. Well, that becomes really difficult for a proprietary AI, right? You're opening yourself up to a third party, and that's a reason why companies may not want to participate. Then there's some other important items like sandboxes. So in AI, research and development is a critical part of this. If we don't have some content around how sandboxes can be leveraged to test out, you know, whether there are actual harms, I think it could become very challenging. So really, at the end of the day, what I think everybody's voicing some concern

on is we don't want to have any unintended consequences, whether it is IP stifling of innovation. At the same time, I think we could all agree, and most Californians probably do agree, that some safeguards for this pervasive technology are necessary. Wendy Sommer provides data labeling and evaluates models. How is the bill and its contents going to impact you in practice? Well, one of the most critical things about developing responsibly AI, is having some human-centric validation. So I believe that this will ultimately drive forward some additional measures for both human-centric validation quality management systems. So I think there could be an increase ultimately over time. But what we do see is that a lot of companies, including big tech, probably do a little bit more than maybe the public is aware of in terms of trying to provide these safeguards. Now it's really about putting forth a regulation that balances that and really creates some transparency.

So we've seen it happen before. Certainly, whether it was green emission standards set up by California or, you know, laws like GDPR in Europe, where they did set forth a wave of actions that ultimately companies had to comply with. So I wouldn't say that, you know, we're seeing it in droves right now. I do think over time, companies are going to need to think about this because, as I mentioned before, I think the unregulated era is coming to an end.

California Passes Controversial AI Safety Bill

Welcome to the Bloomberg Technology YouTube channel. Subscribe for original series and mini-docs on game-changing technologies, profiles on innovators, live conversations with experts, and breaking technology news & analysis from Bloomberg TV.