Elon Musk wants the companies building the most powerful artificial intelligence systems to start checking each other's work. In a recent interview, the Tesla and SpaceX CEO proposed that frontier AI labs — including OpenAI and Anthropic — submit their models to peer review by rival teams.
What Musk proposed
Musk said the current approach to AI safety relies too heavily on internal testing. He argued that outside evaluation, similar to the peer review process used in academic research, would catch flaws that developers might miss or downplay. The idea is that labs like OpenAI and Anthropic would evaluate each other's systems before they are deployed widely.
Musk did not provide a detailed framework. He did not name specific models or set a timeline. But he framed the proposal as a way to build public trust in AI development, which he described as moving faster than society's ability to manage it.
Why the call matters
Frontier AI models — the kind that can generate text, code, and images with near-human ability — have raised concerns about misuse, bias, and catastrophic risks. Right now, safety testing is mostly done internally by each company. Musk's suggestion would introduce an independent check, though it's unclear how that independence would be guaranteed when the reviewers are also competitors.
Musk himself has a complicated relationship with the field. He co-founded OpenAI in 2015 but left the board in 2018. He later started his own AI company, xAI, which builds the Grok chatbot. That puts him in a position where he's both a competitor and a critic of the labs he wants to see review each other.
Who would be involved
Musk specifically mentioned OpenAI and Anthropic as the kind of labs that should participate. Both are considered frontier AI developers. OpenAI builds GPT-4 and the newer o1 models. Anthropic, founded by former OpenAI employees, builds the Claude family of models. Musk's own xAI was not mentioned in the proposal, but it also develops frontier systems.
The idea of peer review is not entirely new. Some researchers have called for external audits of AI models before release. But Musk's public endorsement gives the concept a high-profile push.
Open questions remain
Musk did not say how peer review would be enforced or who would oversee it. He did not address whether labs would be required to share proprietary details about their models to make review possible. And he did not explain how reviewers would avoid conflicts of interest when evaluating a direct competitor's product.
Neither OpenAI nor Anthropic has publicly responded to the proposal. It is not clear whether either lab would agree to such an arrangement, or whether they would demand reciprocity. Musk's own xAI was not mentioned as a participant, leaving the question of whether he would submit his own models to the same process.
