US to reveal which existing laws already bind AI labs, Pentagon CTO says

Pentagon chief technology officer Emil Michael says the wait is almost over. Within one or two weeks, Washington will identify which rules already on the books govern AI labs in the US, he said. Michael also serves as second-in-command of the government's Super Intelligence Task Force. What he has not done is name the laws in question or explain how anyone would enforce them.

That silence keeps AI developers guessing. Michael's stated preference is to clarify existing regulation rather than stand up a new licensing system. His argument: transparency builds public trust, and it might persuade Congress to hold off. Otherwise, he warned, you get 535 legislators writing rules that could kill the industry.

It's a familiar play for this White House. President Trump's June 2 executive order directed agencies to prepare classified assessments of frontier AI models. It floated a voluntary review process that would give authorities up to 30 days of access to covered models before they reach trusted partners, and it explicitly ruled out mandatory pre-clearance.

Europe moved faster, and put dates on it. Since August 2, 2025, the EU AI Act has bound providers of general-purpose AI to keep technical documentation, publish training-data summaries and follow copyright rules. Models judged to carry systemic risk — those trained above 10^25 FLOPs — face extra duties including risk assessment, incident reporting and cybersecurity safeguards. Whether Washington's clarification delivers anything close to that certainty is an open question.

The US isn't drafting alone. A CSIS analysis counts three competing efforts: the US-led Pax Silica coalition with 24 partner countries, the UN Global Dialogue on AI Governance, and China's WAICO, which launched with 29 signatories. No G7 or EU member joined WAICO. CSIS argues that if Washington wants a louder voice in those negotiations, it needs a durable regulatory framework at home first.

Testing access is the next front. An October Brookings briefing found that AI test results increasingly drive procurement and export-control decisions, even as the benchmarks behind them grow less reliable. Cryptopolitan reported that Washington pushed to review frontier models ahead of British evaluators. The stakes keep climbing: the 2026 Singapore Consensus reports frontier-model revenues up more than 400% year over year, BCG found nearly half of companies now generate meaningful value from AI, and Microsoft's research shows adoption remains uneven globally.