Musk Urges US, Chinese Labs to Cross-Test AI Before Release

Elon Musk wants rival AI labs, including "three or four of the leading Chinese companies," to run each other's safety tests before any frontier model ships.

The framework he sketched to CNBC is mechanical, not diplomatic. "All the AI companies have a test harness, a series of tests that you give to any model to see if it's going to build bioweapons or nuclear bombs or be deliberately deceptive," Musk said, arguing that "everyone applying everyone else's test harness is probably the best thing we can do to ensure our safety." Labs would grant rivals API access a week or two ahead of launch so "competitors can keep each other honest." If a competitor concludes a model is unsafe and the developer refuses to fix it, the competitor goes public. "The legal liability would be enormous," he said.

China is not a footnote in the pitch; it is the whole point. Musk called a US-China arrangement "the only thing I can think of that we could probably get all parties, including China, to agree to," proposing "peer review, where the leading AI companies all test each other's models before release."

Anthropic and OpenAI do not come out even. "On balance, I think Anthropic puts more care into their safety than OpenAI," Musk said. On X, he backed Dario Amodei's call to slow the frontier: "Peer review of AI by competitors is the right way to start this off."

The proposal lands in the same week Amodei told CNBC that China is the "toughest dilemma" for any slowdown, and Sam Altman endorsed the same brake-tapping impulse. It is one of several safety-governance stories our tracker has logged this week, alongside OpenAI's endorsement of the FRONTIER Act and an Anthropic co-founder's push for verifiable kill-switch rules.

Musk did not name which Chinese firms. Neither did the reporting.