What happened
On 18 September 2026 Reps. Josh Gottheimer (D-NJ) and Mike Lawler (R-NY) announced two bipartisan House bills: the American AI Security Act, which would require developers of the most powerful AI models to give the NSA full access for a mandatory pre-deployment national-security review (30-day clock with a single 30-day extension) to assess cyberattack or chemical/biological/radiological weapon potential before release; and a companion bill banning federal agencies' use of Chinese-developed AI systems.
Why it matters
This is the first major bipartisan US proposal to impose a mandatory, government-run safety gate on frontier model releases — a direct US federal analogue to the international Call for Control and to state-level RAISE Act rules. If enacted it would create a hard pre-release barrier for the most capable models and a federal procurement ban affecting Chinese model (e.g. DeepSeek-class) deployments in government.
Action needed
Frontier labs should prepare for possible NSA pre-deployment evaluations; government contractors should map any Chinese-AI component exposure ahead of a potential procurement ban.