Editorial illustration for Anthropic launches Fable 5, blocks cybersecurity, biology, chemistry queries
Anthropic launches Fable 5, blocks cybersecurity,...
Anthropic released its new Fable 5 model today, the same day its Mythos Preview period ended. The company immediately blocked the model from answering queries about cybersecurity exploits, biology, and chemistry. Anthropic called those topics too volatile for general use. Fable 5 shares its core architecture with the also-launched Mythos 5, but the restricted version is only available to a small group of cyberdefenders vetted through its Project Glasswing program.
Anthropic Tuesday publicly released Claude Fable 5 , its first "Mythos-class" model that it says surpasses its previous frontier Opus models in overall capabilities. But the model's launch today comes with safeguards designed to prevent it from answering queries on topics like cybersecurity, biology, and chemistry, where the company has publicly worried about its potential impact to "uplift" malicious actors.
The launch frames a specific security trade-off. Anthropic is restricting its most capable model to a pre-approved group, arguing some knowledge is inherently dangerous. That approach leaves the company's Project Glasswing as the gatekeeper for advanced AI access in sensitive fields. We don't yet know how the system will handle novel prompts that skirt its defined prohibitions.
Related Reading
Browse all Anthropic news →
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
Related Topic Trump cracks down on Anthropic after Amazon tip; staff largely foreign
Open Source
Related Topic Anthropic says Claude writes >90% of its code; AI pause button urged
LLMs & Generative AI
From Archives Datadog engineers start AI coding firm Niteshift, backed by Hoffman, Pomel
From Archives Moonshot AI seeks USD 30 billion valuation, plans USD 1‑2 billion fundraise
From Archives Claude Fable declines basic biology queries; Opus 4.8 responds
LLMs & Generative AI
From Archives Four New Specific Techniques to Boost Productivity with Claude Code
LLMs & Generative AI
Common Questions Answered
Why did Anthropic block cybersecurity, biology, and chemistry queries on Fable 5?
Anthropic determined that these three topics are too volatile for general use and pose potential security risks if widely accessible. The company decided to restrict access to these sensitive knowledge areas immediately upon Fable 5's release to prevent misuse of the model's capabilities.
What is the relationship between Fable 5 and Mythos 5?
Fable 5 and Mythos 5 share the same core architecture, as both models were launched by Anthropic on the same day. While they are built on similar foundations, Fable 5 implements additional content restrictions that Mythos 5 does not have.
How does Anthropic's Project Glasswing relate to Fable 5's restricted access model?
Project Glasswing serves as the gatekeeper for advanced AI access in sensitive fields, controlling who can use Fable 5 for restricted topics like cybersecurity, biology, and chemistry. This approach represents Anthropic's strategy of limiting the most capable model to a pre-approved group rather than making it universally available.
What security trade-off does Fable 5's launch demonstrate?
Fable 5's launch illustrates Anthropic's decision to restrict its most capable model to specific users while blocking certain knowledge domains, arguing that some information is inherently dangerous for general distribution. This trade-off prioritizes safety by limiting access rather than making advanced AI capabilities universally available to all users.
Further Reading
- Claude Fable 5 and Claude Mythos 5 — Anthropic
- Anthropic's 'safe' Mythos-class model won't answer questions about cybersecurity or biology — Business Insider
- Anthropic Releases Claude Fable 5, Its Most Powerful AI Yet, With Guardrails for Cybersecurity and Bio — The Hacker News
- Claude Fable 5: Anthropic Locks Down Cyber and Bio — Aardwolf Security
- Anthropic has launched Claude Fable 5, its most capable model ever made available to the public — AIM Media House
