Anthropic Withholds Its Most Capable Model, Saying 'Mythos' Is Too Dangerous to Release
Anthropic has declined to publicly release its newest frontier model, internally codenamed Mythos, claiming its capabilities cross safety thresholds the company isn't yet comfortable exposing to the open internet. The decision reignites the industry's most consequential debate: who decides when AI is too powerful?
Anthropic has made the extraordinary decision to withhold its latest frontier model from public release, telling reporters and researchers that the system — internally referred to as Mythos — exceeds capability thresholds the company considers unsafe for general deployment. As @DonnaG70580 noted, Anthropic explicitly stated that its "newest AI model is too powerful to release to the public," a framing that is as remarkable for its candor as it is for its commercial implications.
The move immediately drew scrutiny. @TechCrunch posed the question that many in the industry are asking: "Is Anthropic limiting the release of Mythos to protect the internet — or Anthropic?" It's a fair question. Withholding a model on safety grounds is one of the few moves that simultaneously burnishes a lab's safety credentials, generates enormous media attention, and avoids the risk of a public benchmark embarrassment if the model underperforms in the wild. Anthropic gets to claim it built something extraordinary without ever having to prove it in an uncontrolled environment.
Get our free daily newsletter
Get this article free — plus the lead story every day — delivered to your inbox.
Want every article and the full archive? Upgrade anytime.
No spam. Unsubscribe anytime.