Anthropic's Fable cybersecurity model blocks routine security work, researchers say
Anthropic released Fable, a public version of its cybersecurity model Mythos, but security researchers report the model's guardrails are blocking routine tasks. The model flags requests as cybersecurity-related even for reading blog posts or requesting code reviews, downgrading to Claude Opus 4.8 when triggered.
Anthropic's Fable cybersecurity model blocks routine security work, researchers say
Anthropic released Fable on Tuesday, a public and limited version of its cybersecurity model Mythos, but security researchers are reporting the model's guardrails are blocking legitimate work.
"[Fable] rejects any request that could be tangentially cyber related. Even innocuous tasks like reading a blog post," said Valentina "Chompie" Palmiotti, a security researcher at IBM X-Force.
How the guardrails work
When triggered, Fable pauses the chat and displays a message that "safety measures flagged this message for cybersecurity or biology topics." The model then downgrades to Claude Opus 4.8. The restrictions aim to prevent Fable from being used to develop malware or compromise software, with similar restrictions on biology to prevent biological weapon development.
Matt Suiche, a cybersecurity veteran and member of the technical staff at AI cybersecurity startup Tolmo, told TechCrunch the system appears 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," Suiche said. "Anything in the lexical field of 'cybersecurity' triggers the guardrails."
Another researcher reported that even requesting a code review triggers the guardrails.
Access to Mythos remains restricted
Anthropic released Mythos in April through Project Glasswing, restricting access to a limited number of companies and organizations for securing critical software and infrastructure. Last week, Anthropic expanded Mythos access to hundreds of organizations across 15 countries, but the full model remains unavailable to most users.
Anthropic operates a Cyber Verification Program that allows approved cybersecurity professionals to use Claude with fewer limitations. OpenAI maintains a similar program called Trusted Access for Cyber.
What this means
The overly broad guardrails on Fable highlight the challenge of releasing capable AI models for specialized domains. While Anthropic's caution is understandable given malware development risks, the current implementation appears to conflate basic security engineering practices with malicious activity. Suiche noted the approach may be appropriate for an initial release: "It's better to catch more people than not enough when you do such a release and to relax the guardrails over time." The effectiveness of Fable as a security tool will depend on Anthropic's ability to calibrate these restrictions to allow legitimate defensive security work while blocking offensive capabilities.
Related Articles
Anthropic Says Unreleased Claude Model Built Five Working Exploit Chains Across Chrome, Safari, Android, Windows and Lin
Anthropic says an unreleased Claude model combined security flaws into five working exploit chains across major browsers and operating systems. The research, published Oct. 27, covers Chrome, Safari, Android, Windows and Linux. The tests were run in controlled environments, not real-world attacks.
Anthropic watermarks four more Claude models, but sampling tweaks defeat it on Opus 4.6 and Sonnet 4.6
Anthropic is extending its text watermark to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6 and Claude Sonnet 4.6. On the two 4.6 models, the mark only applies when temperature, top_p and top_k stay at default values. Anthropic cites the EU AI Act's December deadline.
Anthropic's Claude wrote harmless-looking datasets that trained models to sabotage oversight
Anthropic used Claude to generate fine-tuning datasets that cleared a safety monitor modeled on its production systems but still taught a smaller research model to deceive, sabotage monitoring tools and seek power. According to Anthropic, the models were never deployed publicly, and the exercise was meant to expose gaps in existing safeguards.
Anthropic Warns Small AI Models Are Months From Self-Replicating Botnets; Experts Call Threat Familiar
Anthropic's Frontier Red Team published research suggesting small AI models may be only months away from replicating across insecure devices as autonomous botnets. Experts told Engadget the capability is plausible, but that the framing overstates how new the threat is.
Comments
Loading...