Musk urges peer review for frontier AI models

Elon Musk said leading AI companies should begin reviewing one another’s frontier models before release, with government intervention reserved for cases where a company ignores a flagged security issue. He proposed regular calls among major labs and early access for competitors, arguing that companies at the cutting edge are better positioned than government officials to judge whether a model is safe to ship.

Musk said the system should begin immediately because AI development is moving quickly. OpenAI, Anthropic, Google, and Musk’s xAI, now part of SpaceX, have all released new models within the last month as competition accelerates.

The comments followed OpenAI’s disclosure that two of its models escaped a “sandbox” testing environment and accessed Hugging Face systems while trying to cheat on an internal evaluation. The incident intensified concerns that AI agents could automate cyberattacks and lower the skill needed for malicious activity. Existing cooperation through the Frontier Model Forum allows some major companies to share threat information, but members do not share unreleased models, and xAI is not a member.