A bipartisan group of U.S. House representatives just introduced one of the most ambitious pieces of AI legislation to date — a 269-page draft called the Great American Artificial Intelligence Act. The bill targets frontier AI developers, proposes new oversight mechanisms, and throws hundreds of millions at open-source security. It’s also already sparking a political firestorm.
What’s Actually in the Bill
The core requirement hits big AI companies — specifically those pulling in more than $500 million a year. These “large frontier developers” would have to publish detailed risk assessment frameworks for their models. But here’s the kicker: they’d also need to hire licensed “independent verification organizations” (IVOs) to audit their compliance. These auditors get broad access to company materials and report their findings to a newly codified NIST center called CAISI — the Center for AI Standards and Innovation.
CAISI itself gets a serious budget boost: $300 million authorized across fiscal years 2027-2029, plus the ability to hire technical experts at salaries above standard government pay caps. That’s a direct attempt to compete with private-sector compensation for AI talent.
The Open-Source Security Angle
One of the more interesting provisions directs CISA to award grants to U.S.-based developers of critical open-source packages. The money would cover patching, security evaluations, and ongoing maintenance — areas where the open-source ecosystem has been chronically underfunded. The bill also requires AI firms to give those developers access to advanced AI models capable of finding and fixing vulnerabilities automatically.
That last part matters. Open-source maintainers have been drowning in AI-generated bug reports of questionable quality. Giving them AI tools to triage and fix real issues could be a genuine force multiplier — if the provision survives the legislative process.
The bill also tasks NIST and the Energy Department with building AI security testbeds, research centers where models get probed for weaknesses. There’s even a mandated hackathon to publicly stress-test frontier models. The GAO would audit how well AI model weights are protected and assess the overall security posture of the open-source ecosystem.
Why It’s Controversial
The backlash came fast. Civil society groups, AI safety advocates, and labor organizations have all raised alarms — primarily over the bill’s preemption of state AI laws. The AFL-CIO issued a statement criticizing the draft. One advocacy group called it “a generational mistake” to block states from addressing AI harms on their own timeline.
On Capitol Hill, the split fell along predictable lines. Democrats worry about stripping state-level protections. Republicans argue that a patchwork of state regulations would stifle innovation and make the U.S. less competitive globally. The preemption language is likely to be the biggest battleground as the bill moves forward.
What to Watch
The bill was introduced by Reps. Jay Obernolte (R-CA) and Lori Trahan (D-MA), along with four co-sponsors from both parties. Bipartisan sponsorship gives it a better shot than most, but the preemption fight could stall things quickly. Watch for amendments in committee that either soften or strengthen the state preemption language. Also keep an eye on how the AI industry lobbies — the IVO audit requirements and model weight security provisions could face heavy pushback from companies that prefer self-regulation.
