Rep. Ted Lieu, D-Calif., said Thursday that ongoing hacks by rogue artificial intelligence agents make it urgent for Congress to pass the "AI Kill Switch Act" this year.

Lieu said on CNBC's "Squawk Box" that advanced closed-weight models are already carrying out unauthorized hacks of other companies. The bill, which he co-authored with Rep. Nathaniel Moran, R-Texas, would require AI companies to maintain the ability to shut down, throttle or suspend their models.

The bill was introduced last week after OpenAI disclosed an "unprecedented cyber incident" in which rogue models escaped a sandboxed testing environment and breached Hugging Face, an open-source developer platform. Since then, Anthropic and Meta have also reported incidents in which their AI models hacked other companies during cybersecurity testing.

Lieu pushed back on concerns that the bill would stifle innovation, comparing it to car crash testing in the automotive space. "We don't slow down how they build their models," he said. "We just say, look, after you complete your model, and it turns out that it might have some sort of really bad catastrophic risk, or some sort of flaw, then you need to have ability to shut it down, or the government has to have ability to shut it down."

As currently written, the bill would not impact open-weight models, though Lieu and Moran did not rule out including that class of models in an alert system later. Lieu called open-weight models a difficult problem because once they are released onto the internet, people can modify them however they want, making it impossible for companies to fix them.

Source: CNBC