New AI regulation bill introduced in Congress

A bipartisan group of US senators introduced legislation Tuesday that would require federal licensing for the most advanced AI systems, marking the most significant congressional push to regulate the technology to date. The bill would establish a new oversight body within the Commerce Department empowered to audit frontier models before they are released. Companies developing systems exceeding a specified computational threshold would be required to submit safety evaluations and register with the new regulator. Industry reaction was mixed. The bill faces a difficult path through committee amid a crowded legislative calendar.

US lawmakers have introduced a wide-ranging bill that would create federal oversight of the most powerful artificial intelligence systems. The measure follows more than a year of hearings and study by a bipartisan working group. It would require pre-deployment safety testing, incident reporting, and public disclosures of training data provenance. The UK and EU have both moved ahead of the US on AI regulation, and the bill's authors said the legislation aims to close the gap. The White House expressed general support but stopped short of endorsing specific provisions.

Silicon Valley leaders are warning that the newly introduced AI licensing bill could stifle innovation and hand a competitive edge to Chinese rivals. Executives at leading AI labs said the licensing regime would add months to development cycles and impose compliance costs that only the largest firms could absorb. Venture-capital investors expressed concern that the bill would freeze early-stage funding in the sector. The trade group NetChoice said the measure amounts to a 'permission slip' requirement for building software. Supporters counter that the threshold is calibrated to affect only a handful of well-resourced developers.

Republican lawmakers are pushing back against portions of the newly introduced AI bill, warning that broad content-safety provisions could enable government censorship. Critics said the bill's authority for the regulator to demand training-data disclosures could be weaponized against models that produce disfavored viewpoints. Conservative legal scholars argued the licensing regime raises First Amendment issues. Supporters of the bill dismissed the concerns as premature and said the enforcement mechanisms are narrowly scoped. A markup in the relevant committee is scheduled for next month.

The AI oversight bill introduced this week reflects years of work by researchers, whistleblowers, and workers whose jobs have been disrupted by the technology. Members of the Writers Guild, the SAG-AFTRA union, and displaced call-center workers testified in Senate hearings that shaped the bill's disclosure provisions. Former employees of leading AI labs called for mandatory safety testing after describing internal pressures to ship models quickly. Advocates said the introduction of the bill represents 'a rare moment' when public concerns broke through the industry's lobbying power.