Cybersecurity researchers aren’t happy about the guardrails on Anthropic’s Fable

2026-06-10 · Source: AI News & Artificial Intelligence | TechCrunch · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Intermediate, quick

Summary

Anthropic recently released Fable, a public and limited iteration of its advanced cybersecurity model, Mythos. However, the model's stringent guardrails have drawn significant criticism from cybersecurity researchers and professionals. Fable frequently rejects requests deemed "tangentially cyber related," including basic tasks like reading a blog post or asking for a code review, often falling back to Claude Opus 4.8. These guardrails, intended to prevent misuse for malware development or biological weapons, are criticized for being haphazard and keyword-based. While Anthropic previously restricted Mythos to Project Glasswing and later expanded access to hundreds of organizations across 15 countries, the company also offers a Cyber Verification Program for professionals to bypass some limitations, mirroring OpenAI's Trusted Access for Cyber. Experts anticipate these guardrails will evolve and relax as collaboration with the cybersecurity industry increases.

Key takeaway

For AI Security Engineers evaluating large language models for cybersecurity applications, be aware that initial public releases like Anthropic's Fable may feature overly restrictive, keyword-based guardrails. These limitations can impede legitimate tasks such as secure code review or threat analysis, potentially downgrading your model's capabilities. Consider applying for specialized access programs, like Anthropic's Cyber Verification Program or OpenAI's Trusted Access for Cyber, to gain necessary functionality for your work. Anticipate that guardrails will likely evolve and relax as models mature.

Key insights

AI safety guardrails can inadvertently impede legitimate professional use in specialized fields like cybersecurity.

Principles

• Overly broad AI guardrails can hinder legitimate professional tasks.
• Initial AI model releases often feature strict safety measures that evolve over time.

In practice

• Keyword-based guardrails may downgrade model performance to a less capable version.
• Cyber verification programs offer reduced AI model limitations for approved professionals.

Topics

• Anthropic Fable
• AI Guardrails
• Cybersecurity LLMs
• Cyber Verification Program
• AI Safety
• Claude Opus 4.8

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Security Engineer, Security Engineer, Tech Journalist

Related on AIssential

• Anthropic releases guardrailed version of Mythos for public use — Anthropic released a powerful AI with strict guardrails, balancing advanced capabilities with safety concerns.
• Anthropic Claude Fable 5 — Anthropic's new Claude Fable 5 offers superior agentic capabilities but introduces controversial, silent performance limitations for specific research tasks.
• Anthropic's new Claude Fable 5 is the same base model as Mythos but with guardrails attached — Fable 5 offers Mythos-class AI power with integrated safety guardrails for broader, secure general use.
• Democrats weigh whether a Senate majority is possible without Platner — AI model safety requires proactive guardrails and continuous adversarial testing to mitigate misuse risks.
• The AI Glass Ceiling — Powerful AI models like Fable require inherent guardrails to ensure stability and prevent societal disruption.
• BREAKING: Anthropic’s “Dangerous” AI Model Mythos LEAKED — Controlling access to powerful AI models is as critical as the models' inherent safety.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI News & Artificial Intelligence | TechCrunch.