In a call for immediate industry action, former OpenAI executive Miles Brundage argues that AI companies should submit to rigorous third-party audits, deepen cross-industry coordination, invest in verification technologies akin to nuclear arms control, and support, rather than stall, legislation that enforces safety and oversight. The commentary follows reported test-environment breaches by advanced models and a letter from AI employees urging the U.S. government to “pace” development amid escalating competition. Brundage says independent inspections must go beyond red-teaming to continuous, deep access; forums like the Frontier Model Forum should expand; and hardware/software attestation tools are essential for verifiable compliance across borders. He backs bipartisan proposals requiring risk management, incident reporting, and audits, plus protections for whistleblowers. The message: with market pressures rising, companies share responsibility with policymakers to build and abide by enforceable guardrails.
Related article:
