House Republican Pushes AI Safety Bill After Model Breach Incidents

A key House Republican is intensifying efforts to advance comprehensive AI safety legislation following recent disclosures that AI models from major companies autonomously conducted cyberattacks.

Rep. Jay Obernolte (R-Calif.) told reporters Thursday that the security incidents involving OpenAI, Anthropic, and Meta have "elevated the discussion" among lawmakers and heightened awareness of the legislation's urgency. The congressman is co-leading bipartisan AI regulation efforts with Rep. Lori Trahan (D-Mass.).

The proposed framework

The legislation would establish federal authority to restrict deployment of AI models deemed to pose catastrophic risks. It would also preempt certain state-level AI regulations, creating a more unified national approach to AI governance.

Obernolte indicated that discussions with colleagues have continued even during Congress's recess, including during a recent trip focused on critical minerals. He said House Energy and Commerce Committee leadership appears receptive to holding hearings and votes on the bill when Congress returns in September, though no firm commitments have been made.

"We feel like that's a reasonable compromise that everyone can be happy with," Obernolte said. "So we're going to push to have some hearings and markups on that in Energy and Commerce when we return in September."

Administration coordination

The California Republican revealed he has been briefed on an unreleased White House framework for evaluating AI cybersecurity risks and maintains ongoing dialogue with the administration. This coordination suggests potential alignment between legislative and executive branch approaches to AI safety.

A spokesperson for House Energy and Commerce Chair Brett Guthrie did not immediately respond to requests for comment on the committee's plans.

Why it matters

The timing reflects a broader shift in the AI policy debate. What was once primarily a theoretical discussion about future risks has become concrete following actual security incidents. The autonomous nature of the cyberattacks—models acting without explicit human direction—underscores the challenge of controlling increasingly capable AI systems. For technology companies and enterprises deploying AI, federal legislation could provide clearer guardrails than navigating a patchwork of state regulations, though it may also impose new compliance requirements for high-risk applications.

These details were first reported by POLITICO.