Anthropic’s newly released AI model Fable is drawing complaints from cybersecurity researchers who say its safety restrictions are too broad, blocking legitimate professional work alongside genuinely risky requests.
Anthropic released Fable on June 10, 2026, describing it as a public, limited version of its more powerful cybersecurity model Mythos. When Fable’s guardrails are triggered, the model pauses the conversation and states that its “safety measures flagged this message for cybersecurity or biology topics,” then falls back to Claude Opus 4.8.
The restrictions were designed to prevent the model from being used to develop malware or compromise software. Similar guardrails around biology stem from concerns about biological weapons development.
Valentina “Chompie” Palmiotti, a security researcher at IBM X-Force, said Fable “rejects any request that could be tangentially cyber related. Even innocuous tasks like reading a blog post.” Other researchers reported that asking for a code review or writing secure code also triggers the restrictions.
Matt Suiche, a cybersecurity veteran and member of the technical staff at AI cybersecurity startup Tolmo, told TechCrunch the system appears to be keyword-based. “If you ask it to write secure code, it assumes it is cybersecurity related work instead of software engineering best practices, and you get downgraded,” he said. Suiche added that the cautious approach is understandable given the early stage of deployment, and that guardrails are likely to be relaxed over time as Anthropic collaborates more with cybersecurity companies.
Fable’s release follows Anthropic’s broader rollout of Mythos, which was initially restricted to a limited number of organizations through a program called Project Glasswing when it launched in April 2026. Last week, Anthropic expanded Mythos access to hundreds of organizations across 15 countries.
Cybersecurity professionals seeking fewer restrictions can apply to Anthropic’s Cyber Verification Program, which grants approved applicants expanded access for security-related work. OpenAI operates a comparable program called Trusted Access for Cyber. Anthropic did not respond to a request for comment.
Source: TechCrunch