Anthropic holds Mythos in lab over biosecurity risks
Anthropic disclosed that its most capable upcoming model — internally code-named Mythos — has been held back from any external API release after the company's safety evals flagged uplift potential in cyber and biosecurity domains.
Anthropic confirmed the deferred release in a brief statement coupled with a longer internal-evals document. The model is described as substantially stronger than Opus 4.7 on offensive cyber and "dual-use biology" red-team prompts, with mitigations still in active development.
The disclosure prompted a notable response from the Trump administration, with the President publicly stating "there should be regulations on AI" — a meaningful shift in rhetoric from the December 2025 deregulatory executive order.