Elon Musk urged leading AI labs and several Chinese firms to run each other’s models through independent safety tests, the CNBC report said Monday after his comments at the All-In Summit.
Speaking at the All-In Summit in Los Angeles, Musk proposed that xAI, OpenAI, Anthropic, Google, Meta and “three or four of the leading Chinese companies” let rivals run a “test harness” on their frontier models to flag safety issues, according to CNBC.
Musk framed the idea as a form of peer review — “instead of grading your own homework, you would at least have competitors grading your homework and raising the alarm if they see concerns,” CNBC quoted him as saying.
The CNBC article notes Musk acknowledged the plan is imperfect and that rival labs had not agreed to the proposal, while Anthropic’s CEO Dario Amodei has suggested third-party “embedded evaluators” as an alternative safety check.
Industry leaders have recently called for slowing frontier-model development, a debate that CNBC linked to a wider set of warnings from executives at Anthropic and OpenAI and to public pushback from political figures, including President Trump.
Analysts and safety researchers have offered different solutions: CNBC reported Amodei’s essay proposing a coordinated slowdown and third-party assessments, while Musk suggested cross-testing could increase the odds of spotting dangerous behaviour before public release.
Why it matters: peer review between competing labs would be an unusual step in a fast-moving, commercially competitive sector and would test how companies balance safety with intellectual-property and national-security concerns, CNBC said.
Sources
- CNBC — reporting of Musk’s All-In Summit remarks, the ‘‘test harness’’ proposal, and context about industry calls to slow model development.
- Quartz — independent coverage noting Musk’s call for rival labs to safety-test each other’s models and that rivals have not agreed to the plan.
