The Senate returns this week to draft the first statutory framework governing the release of frontier A.I. models, following a summer in which the executive branch demonstrated, twice, that it can already stop a model from shipping without one. What began as an emergency posture is now being retrofitted into law.

The precipitating episode is documented in a Congressional Research Service brief, IF13217. On June 12, the Commerce Department invoked national security export controls to require Anthropic to disable its Fable 5 and Mythos 5 models. According to Anthropic, the directive arrived at 5:21 p.m. Eastern time and applied to “any foreign national, whether inside or outside of the United States,” a scope that extended to the company’s own non-citizen employees. The models stayed offline for 18 days. Commerce lifted the restrictions on June 30.

Anthropic pushed back publicly. “We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people,” the company said, noting that comparable capabilities could be elicited from publicly available systems, including OpenAI’s GPT-5.5. It also endorsed, in the same statement, a “statutory process that is transparent, fair, clear, and grounded in technical facts.” Translation: legislate the constraint so it stops arriving at 5:21 on a Friday.

Days after the Anthropic models came back online, OpenAI told TechCrunch it was limiting the GPT-5.6 lineup, Sol, Terra, and Luna, to a “small group of trusted partners,” calling the move a “short-term step” while it worked with the administration on a “repeatable process for future model releases.”

The policy scaffolding is already partly built. Executive Order 14409, signed June 2, directed the creation of a voluntary federal evaluation framework, housed at the Center for A.I. Standards and Innovation inside Commerce. That voluntary posture is what the Senate is now trying to convert. According to The Washington Post, Majority Leader John Thune and Senator Amy Klobuchar are examining a “duty of care” model grounded in developer self-testing, while Senator Maria Cantwell is weighing a different approach. Policy specialists are debating whether to move the Center out of Commerce entirely, an implicit rebuke of how the June episode was handled.

Peter Girnus, a cybersecurity researcher, put the industry’s position in unsentimental terms to Fortune: “If you describe your product as a munition in every press release, eventually a government takes you at your word.” The labs are now asking Congress to draw the line they spent three years insisting existed.

Sources