Anthropic releases Claude Fable 5 with guardrails on high-risk topics

Anthropic is putting its most sensitive model family in public view, but only after building a wall around the parts of it it fears most. Claude Fable 5 is the company’s first Mythos-class model available to the public, and Anthropic said it will block or reroute requests in high-risk areas such as cybersecurity, biology, chemistry and distillation.

The release underscores the balance Anthropic is trying to strike between reach and restraint. Fable 5 uses safeguards that send narrow high-risk prompts to Claude Opus 4.8, Anthropic’s latest generally available model before this launch. The company said internal and external red-team testing found no known universal jailbreak technique, a sign that it is trying to commercialize frontier capability without repeating the backlash that has followed more aggressive rollouts elsewhere in the industry.

Fable 5 is built on the same underlying Mythos model Anthropic previewed on April 7, when it first described the system as unusually capable at cyber tasks. At that point, Anthropic kept access tightly limited through Project Glasswing, a defensive cybersecurity initiative launched the same day with partners including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA and Palo Alto Networks. Anthropic later extended access to more than 40 additional organizations and said it committed up to $100 million in usage credits and $4 million in donations to open-source security groups.

The company has leaned hard on the security case for Mythos. Anthropic said the model had been used to scan more than 1,000 open-source projects and that it began using an early snapshot in February 2026 to look for vulnerabilities in open-source software. That work helped fuel broader conversations with governments, software maintainers and researchers as concerns grew over offensive cyber use.

Those concerns intensified in May, when CEO Dario Amodei warned that there was only a six- to 12-month window to patch tens of thousands of vulnerabilities uncovered by Mythos before Chinese AI systems caught up. The public release of Fable 5 arrives alongside Opus 4.8 and comes just days after Anthropic confidentially submitted a draft S-1 to the SEC and raised $65 billion at a $965 billion post-money valuation on May 28.

For Anthropic, the strategy is clear: widen access, but do it with explicit guardrails. Whether that distinction proves meaningful will shape how safely the next generation of powerful AI systems reaches the public.