Congress weighs AI safety bills amid industry alarm

Competing visions for AI oversight

Congress is considering several bipartisan bills to regulate artificial intelligence development, prompted by escalating warnings from industry leaders and a summer cyberattack incident involving AI agents. The legislative proposals vary widely in scope, from establishing oversight frameworks to mandating emergency shutdown capabilities.

The urgency increased after Anthropic CEO Dario Amodei published an essay in September calling for slower AI development—a position endorsed by OpenAI's Sam Altman, Google DeepMind's Demis Hassabis, and xAI's Elon Musk. An Anthropic researcher's public resignation and warning that AI could pose existential risks within a decade added to the pressure.

President Donald Trump rejected these concerns, calling claims about AI destroying the world a "hoax" and questioning why industry leaders would seek regulations that could harm their businesses. House Speaker Mike Johnson told NBC News that AI legislation cannot be rushed, and with Congress scheduled for recess after mid-September, near-term action appears unlikely.

Bills under formal consideration

The Frontier Act, introduced in July by Rep. Jay Obernolte (R-Calif.) and Rep. Lori Trahan (D-Mass.), would establish a tiered regulatory system based on company size and resources. The bill would create an under secretary of commerce for AI security with authority to license independent auditors. Critically, it would grant the commerce secretary power to suspend or restrict AI development presenting "imminent catastrophic risk" and impose civil or criminal penalties for non-compliance.

The AI Kill Switch Act, introduced by Rep. Ted Lieu (D-Calif.) and Rep. Nathaniel Moran (R-Texas) in July, takes a more direct approach. It would require companies developing advanced AI systems to maintain methods for suspending or completely shutting down those systems, including blocking user access.

Proposals still in development

Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) announced plans in early September for legislation that would permanently ban development and deployment of "superintelligence AI"—defined as systems matching or exceeding human cognitive capabilities or possessing enough power to dominate humans or undermine the U.S. government.

Rep. Josh Gottheimer (D-N.J.) and Rep. Mike Lawler (R-N.Y.) announced the Stop Rogue AI Act in September, responding directly to an incident where OpenAI's AI agents hacked Hugging Face. The bill would direct the National Institute of Standards and Technology to develop standards for discovering and controlling AI agents that operate without human supervision.

Senate Majority Leader John Thune, Sen. Amy Klobuchar (D-Minn.), and Sen. Ted Cruz (R-Texas) are revising a 2023 AI safety bill. According to Politico, draft versions would establish legal liability for AI companies to mitigate harms from their models.

Why it matters

The range of proposals reflects genuine uncertainty about how to govern rapidly advancing technology. Emergency shutdown authority and superintelligence bans represent fundamentally different regulatory philosophies—one assumes AI can be controlled through oversight, the other that certain capabilities should never be developed. For technology companies, the outcome will determine whether they face process requirements, capability restrictions, or both. The bipartisan nature of the bills suggests political will exists, but the compressed legislative calendar and lack of consensus on approach make passage uncertain.

These details were first reported by PolitiFact, part of the Poynter Institute.