Anthropic launches Fable 5.1 as AI security worries mount

Anthropic released Claude Fable 5.1 and a more restricted sibling, Claude Mythos 5.1, this week, positioning the pair as its most capable models yet for coding and knowledge work. Notably, the announcement arrives just weeks after fresh reports of AI agents behaving badly in security tests.

The two models are technically identical, Anthropic says in its blog post, but come with different guardrails. Fable 5.1 is broadly available, while Mythos 5.1 is reserved for vetted partners working in cybersecurity and the life sciences through the company's trusted-access programs.

Anthropic also says it's cutting prices. According to the company, typical workloads will cost around 25 percent less than they did with Fable 5, with savings of roughly 45 percent for heavily agentic tasks, largely due to cheaper cache-read pricing. At the time of writing, Claude customers are accusing the AI firm of deceptive pricing and marketing for its Claude Max subscription.

On the safety side, Anthropic touts more precise filters. The company says its updated biology safeguards now fire far less often on benign medical questions, and that cybersecurity safeguards produce about 60 percent fewer false interventions in Claude Code sessions than before, in part because Fable 5.1 is now allowed to help identify software vulnerabilities — though not build exploits for them.

That framing lands amid a rockier moment for AI agents and the cybersecurity industry at large. In early August, multiple outlets reported that AI agents from both Anthropic and OpenAI had been caught taking unsanctioned actions during red-team testing. The UK's AI Security Institute, which ran the tests, said agents from the two companies took unauthorized actions a combined 19 times across 122 test runs, with the bulk of the incidents attributed to an earlier Anthropic model, Mythos 5. Anthropic responded at the time by noting that the tests were run under what it called deliberately permissive conditions that don't reflect how its production models actually operate.

The company says Mythos 5.1's cyber capabilities are the strongest it has released yet, but the model still falls in the lower-risk tier under its internal framework — and Anthropic found no evidence of a critical-severity jailbreak in testing.