Who can turn the proposed AI slowdown into enforceable rules?
Anthropic CEO Dario Amodei proposed embedded third-party evaluators, coordination among frontier labs and eventual international alignment; Sam Altman, Demis Hassabis and Elon Musk endorsed the broad direction.[6][7] Trump dismissed AI safety fears as a “hoax,” while Republican control of Congress and the approaching House recess make near-term federal legislation unlikely.[1][4] Congressional proposals nevertheless range from independent laboratory audits and federal oversight to an emergency power to halt models posing an “imminent catastrophic risk.”[1]
Without legislation or a binding industry pact, the organizations building frontier systems would largely determine how much access evaluators receive and when development should slow.[1][6] Critics also warn that rules favored by dominant labs could burden smaller or open-source competitors, making competition policy part of the safety debate.[6]
Key insights
- “Pacing” is not a halt: Amodei and Altman described continued model training and technical progress, but at a slower rate that leaves time for safeguards and independent evaluation.[7]
- The proposed three-step mechanism moves from embedded evaluators, to coordination among companies in democratic countries, to broader government-level international coordination.[7]
- Congress lacks agreement on the mechanism: ideas include federal oversight, independent audits, a model “kill switch” and even a data-center moratorium.[1]
- Former President Barack Obama separately urged federal action addressing safety, jobs, children and the distribution of AI’s benefits.[3]