It says new safeguards make it possible to release a Mythos-class model it previously said was too risky to make public.
Back in April, Anthropic unveiled its Claude Mythos AI model that it said was too powerful to publicly release.