Story perspectives

White House Pushes Voluntary AI Testing, Mulls FDA-Style Safety Order

5/10/2026

Summary

Story summary

Senior White House officials say the administration prefers partnership with AI firms over regulation, expanding voluntary testing agreements with Google DeepMind, Microsoft, xAI and others and weighing safety measures.
The White House is studying an order that would require safety testing of AI models on release, likening it to FDA drug reviews.
NEC Director Kevin Hassett cited Anthropic’s Mythos model, which finds decades-old software vulnerabilities, as prompting the safety review.

History

Historical Context

Today's calls for AI safety testing echo the 1962 Kefauver-Harris amendment that forced the FDA to require pre-market safety trials, a shift rooted in earlier consumer-protection laws such as the 1906 Pure Food and Drug Act and the 1938 Federal Food, Drug, and Cosmetic Act. The precedent of federal oversight was later reinforced by President Richard Nixon’s 1970 Controlled Substances Act.

1 of 2

Story summary

Senior White House officials say the administration prefers partnership with AI firms over regulation, expanding voluntary testing agreements with Google DeepMind, Microsoft, xAI and others and weighing safety measures.
The White House is studying an order that would require safety testing of AI models on release, likening it to FDA drug reviews.
NEC Director Kevin Hassett cited Anthropic’s Mythos model, which finds decades-old software vulnerabilities, as prompting the safety review.

1 / 2