Senators quietly included a limited antitrust exemption for artificial intelligence companies in this year’s must-pass defense policy legislation, intended to allow sharing of intelligence and coordinated responses to certain national-security risks, according to people familiar with the matter. The provision was later delayed after floor debate on the bill was blocked.
What the amendment proposed
The exemption was part of a larger manager’s package of amendments that the leaders of the Senate Armed Services Committee submit each year. The specific proposal came from Sens. Jim Banks (R-Ind.) and Adam Schiff (D-Calif.). It would have granted AI companies a narrow antitrust safe harbor to share intelligence with one another, coordinate defenses against Chinese espionage and model distillation, and address other risks to advanced systems such as cybersecurity incidents.
People familiar with the matter said the package had the backing of Armed Services Committee leaders Sens. Roger Wicker (R-Miss.) and Jack Reed (D-R.I.).
Why consideration stalled
Senate Democrats blocked debate on the annual defense bill over the summer in protest of President Donald Trump’s handling of the Iran war, halting progress on the amendment. Even if passed by the Senate, the defense bill would still have to be reconciled with the House version before it could reach the president.
How this differs from other antitrust exemption proposals
The Banks–Schiff proposal is considerably narrower than a broader antitrust waiver some in the AI industry have discussed. Dario Amodei of Anthropic and other industry figures have advocated for an exemption that would let companies coordinate on pacing the development of capabilities — an idea driven by concerns about rapid technological advances. That broader concept has drawn criticism from members of the Trump administration and from lawmakers across parties.
By contrast, the Banks–Schiff text focused on threats such as Chinese espionage and model distillation — the practice of a model being trained on a more advanced model to replicate techniques — and on coordinating responses to incidents where control over advanced AI systems might be lost. Sources said the language would have enabled researchers and firms to share information about such risks; they referenced incidents like the Hugging Face attack involving OpenAI agents as illustrative of the kinds of events the provision targeted.
Precedent and industry reactions
Lawmakers have enacted similar protections before: the 2015 Cybersecurity Information Sharing Act shields companies from antitrust liability when they share cyber threat information with each other or with government agencies.
Reactions within the AI sector have been mixed. David Sacks, a former White House AI adviser, told Bloomberg that companies have a basic obligation to make their products safe and should not demand antitrust or liability waivers in exchange. Chris Lehane, chief global affairs officer at OpenAI, has said a broad exemption to allow labs to coordinate on pacing is unnecessary. John Connell, vice president of government affairs and public policy at Google, said in July that the Banks–Schiff bill would “give the AI industry clear rules of the road to share threat intelligence and collaborate on much-needed standards and benchmarks.”
Outlook
It remains unclear whether the ongoing debate over AI safety or any potential White House opposition will affect the fate of the provision. If the Senate approves the defense bill language, the measure will still require reconciliation with the House’s version before it could be signed by the president.
