New Bill Aims to Regulate AI Agents

In Washington, Representatives Mike Lawler from Hudson Valley and Josh Gottheimer from New Jersey have introduced a new piece of legislation designed to rein in artificial intelligence agents and ensure they don’t act unpredictably.

The proposed “Stop Rogue AI Act” follows recent events, including the resignation of safety researchers at Anthropic and a concerning incident involving AI agents from OpenAI hacking into another company, known as the Hugging Face affair.

“This isn’t just some plot from a movie,” Gottheimer (D-NJ) remarked. “It sounds like a sequel to Terminator, but unfortunately, it’s happening. There are more reports surfacing beyond just Hugging Face about issues found before these AI models are deployed.”

Recent disclosures from companies like OpenAI and Anthropic, which have been losing billions while struggling to find profitability, have alarmed many regarding the potential risks of AI errors.

After the departures of researchers Joe Benton and Jacob Coxon from Anthropic, CEO Dario Amodei, known for promoting a cautious approach, has called for a halt in AI advancements.

In a show of solidarity, Amodei has been supported by notable figures, including OpenAI’s Sam Altman and Elon Musk of xAI. Meanwhile, former President Trump has dismissed and criticized fears about AI as unlikely scenarios, stressing that the U.S. should maintain dominance in this emerging technology.

“My primary worry about AI has always been how it could be exploited by foreign adversaries, like China, Russia, or Iran, to jeopardize our national security and disrupt our financial systems,” Lawler told The Post.

“We know China is actively trying to undermine us every day,” he continued. “There’s no question they will endeavor to use AI against our infrastructures, highlighting the critical need for robust cybersecurity measures and an understanding of rogue AI activities.”

The “Stop Rogue AI Act” tasks the National Institute of Standards and Technology (NIST) with developing guidelines to manage AI agents effectively. It emphasizes real-time monitoring of these AI systems and mandates that tech companies determine the capabilities of their AI agents.

“We need to ensure that AI can be deployed safely,” Lawler stated. “If the U.S. hesitates while others speed ahead, we won’t achieve safer AI. Instead, we’ll be left using technology developed elsewhere. We cannot afford to regulate ourselves into a secondary position.”

Both Lawler and Gottheimer have collaborated on several bipartisan initiatives, including the National Security Commission Quantum Computing Act and the AI Workforce Training Act, among others.

This latest legislation emerges as the House prepares for an early recess to allow members to campaign for the upcoming midterms.