On Tuesday, Anthropic launched a new “Mythos-class” model, Fable 5, describing it as being more powerful than any other model the company had ever made generally available. It’s a significantly tamer version of Mythos, the behemoth model which Anthropic announced in April but has thus far withheld from the public due to its alleged sophistication at finding and exploiting cybersecurity vulnerabilities. If users ask Fable 5 about potentially sensitive subjects like cybersecurity, biology, chemistry, the company said, it will respectfully decline to respond, and instead automatically revert to an earlier model, Opus 4.8. Almost immediately, users started complaining that Fable’s safety guardrails are too sensitive, refusing to answer sometimes comically harmless requests. Ask the model a question that you might see on a third-grader’s biology homework, for example, and it’s liable to take that as a red flag, as if you’re planning to build some biological doomsday weapon. You're not even allowed to ask Fable about basic biology questions, let alone anything that could potentially be dangerous. pic.twitter.com/FOlGpPJqsB — Crémieux (@cremieuxrecueil) June 9, 2026 This hypersensitivity was built in by design, and Anthropic gave ample warning that this would lead a portion of benign prompts to be flagged as dangerous. “To release the model both safely and quickly, we’ve tuned these safeguards conservatively,” the company wrote in its blog post announcing the release of Fable 5, along with another, more powerful model with limited availability called Mythos 5. “With more capable models arriving in the coming months, we’re working to improve our safeguards and reduce false positives as quickly as we can.”