Anthropic releases Mythos-class Fable 5 with safeguards for cyber risks

Anthropic expanded access to its Mythos model while implementing safeguards that reroute certain sensitive categories—cybersecurity, biology, chemistry, and model-distillation queries—to a less-capable model (Claude Opus 4.8). Early testing by security researchers suggests these classifiers may be more conservative than claimed, automatically redirecting routine cybersecurity tasks and potentially broadening the scope of blocked requests.