Sam Altman Calls Anthropic's Mythos a 'Bomb Shelter' Sales Pitch

OpenAI CEO Sam Altman has publicly accused Anthropic of using safety fears as a marketing strategy, responding to the company's decision to restrict its new cybersecurity AI model Mythos to a small group of enterprise customers.

Anthropic released Mythos as a model it considered too dangerous for public access, citing concerns about criminal misuse. Responding during a podcast appearance reported by TechCrunch, Altman offered a pointed reframe: "It is clearly incredible marketing to say, 'We have built a bomb, we will sell you a bomb shelter for $100 million.'" He also argued that framing AI as uniquely dangerous concentrates powerful systems among wealthy organizations while effectively locking out smaller players and the general public.

Anthropic did not respond to Altman's comments. The company has not published independent technical assessments to support its restricted-release decision, so the claim that Mythos poses singular dangers remains unverified by outside parties.

Altman's criticism is pointed, but the territory is complicated. OpenAI has also used safety language to justify its own decisions about what to release and when, and the line between genuine precaution and strategic positioning is rarely clear from the outside. What the exchange makes plain is a real disagreement inside the industry about who should control the most capable AI systems, and about how risks get communicated to the public.

As AI labs build increasingly specialized tools for high-stakes domains like cybersecurity and intelligence analysis, decisions about access will face more scrutiny. Anthropic and OpenAI are already competing aggressively on lobbying, and the Mythos dispute suggests that competition is extending into how each company positions its safety credentials.