Illinois Makes Strides In Regulating A.I: ‘No Technological System Is Inevitable’

**Sasha Ann Simons** (0:12)
Hey Chicago, I'm Sasha Ann Simons, and this is In the Loop.
And at the end of this podcast, stick around for a rundown of the latest news here in the Chicago area. But first, the rules around artificial intelligence are still being written, literally. As we watch AI tools become more common and sophisticated, Illinois lawmakers are trying to figure out how to protect the public without slowing down innovation. They passed legislation aimed at regulating the tech. It requires AI developers to do more, like disclose risks and better explain what their systems can do.
In this episode of the podcast, we'll talk with an AI expert and a state lawmaker to discuss the challenges of enforcement and whether states can fill the void left by Congress. At the table was Democratic State Representative Daniel Didech of Illinois' 59th district in the Northwest suburbs, and Michael Bennett, Associate Vice Chancellor for Data Science and AI Strategy at the University of Illinois Chicago. This bill is being hailed as a first of its kind AI safety legislation. Now, I gave a brief synopsis there, but we needed more, so I had the representative fill us in.

**Daniel Didech** (1:24)
As you mentioned, artificial intelligence is developing at really an unbelievably rapid pace.
If it continues to develop at the rate it has over the last few years, it will be easily one of the most significant technological innovations in our lifetime and potentially in the history of humanity. And it has lots of positive potentials, right? We think artificial intelligence is going to make us healthier, it's going to make us safer, it's going to make us more productive. But there's a flip side to that. And the flip side is that these systems are becoming so powerful that without safeguards in place, they will have the ability to really cause some really catastrophic harms. And when we say catastrophic harms, we're talking about specifically the possibility that these technologies will be used to develop nuclear, chemical, or biological weapons, or be used to participate in conduct that if it were conducted by a human would constitute murder, theft, or extortion. So we passed in Illinois SB 315, the Artificial Intelligence Safety Measures Act. What that law will do is make sure that the largest and most powerful AI systems will have safeguards in place to make sure that there's a plan to make sure that they're not causing catastrophic risks, that there's third-party independent audits, and that there's both governmental oversight and internal safeguards in place so whistleblowers can report catastrophic risks without risking their jobs or their careers.

**Sasha Ann Simons** (2:57)
Right. So Michael, let's pick up where the representative left off there because I know that you've read through, I understand, the full lengthy legislation. And so on the topic of catastrophic risk, because it's mentioned in there quite a bit, tell me, are there credible risks of AI attacks? And if you can give us examples of what that could look like.

**Michael Bennett** (3:19)
First of all, I should say it's a real pleasure to be here. I love the show.

**Sasha Ann Simons** (3:22)
Thank you.

**Michael Bennett** (3:23)
It's great and it's an honor to be on with the state rep as well. Greetings to you.
And so as far as credible, significant risks go, the companies themselves, the very companies that are responsible for introducing these technologies, many of them at least acknowledge that it's certainly the case. Governments around the world, especially large ones in the West, also acknowledge this after having implemented these technologies in various contexts. And then experts in nonprofit institutions and academic institutions like mine, like the University of Illinois system, also have indicated that the answer to this question is yes. The types of examples have already been eluded to by the state rep. The use of generative artificial intelligence and agentic AI systems in military contexts to make targeting decisions independent of human oversight or even in combination, some people would say, with human oversight is really quite problematic.

**Sasha Ann Simons** (4:27)
So Illinois is now one of a handful of states that are pushing through laws to govern AI. And so there's not this patchwork of different laws across states and municipalities, but as I mentioned, no real federal regulations. Why is that?

**Michael Bennett** (4:42)
I'm happy to speak to that.
One of the main reasons is that this is a very difficult domain to regulate. This is an extremely arcane technological domain. It's grounded in a subject matter that many people are horrified by, namely mathematics and even more precisely linear algebra. It's very difficult to understand, first of all. Secondly, at the federal level in the United States, we have a set of impulses among our leaders that leans towards a laissez-faire approach to the technology, right? The fears that, to the extent that you introduce regulation relatively early on in the life of this technology is not even four years old. To the extent that you do this at such an early stage, you are likely to impede its development and growth. And so there's that kind of fear at the federal level, that concern.