Anthropic launched Claude Fable 5 on June 9, calling it the most powerful AI model the company has made widely available. But the release comes with a catch: some of the model’s most sensitive capabilities are locked behind safety systems.

Fable 5 is the first public model from Anthropic’s Mythos class, the same family behind Claude Mythos 5, its more advanced system for cybersecurity and technical research. Mythos 5 remains restricted to vetted partners through Project Glasswing, while Fable 5 is the version available to general users.

The split shows how Anthropic is trying to square two competing goals: releasing frontier AI systems to the market while limiting access to capabilities it believes could be misused.

Fable 5 uses safety classifiers to detect prompts in sensitive areas including biology, chemistry, cybersecurity, and model distillation. When those systems are triggered, the model can refuse the request or route the user to Claude Opus 4.8, Anthropic’s former flagship model.

That means users may not always be interacting with Fable 5’s full capabilities, even when paying for access to the newer model.

Anthropic has said most sessions are not affected by the safeguards, with more than 95% of Fable 5 usage reportedly continuing without interruption. For everyday use cases such as writing, coding, research, and project management, many users may never notice the restrictions.

The limits are more visible for scientists, security researchers, and developers working in areas Anthropic considers higher risk. A biology question, a chemistry prompt, or a cybersecurity query may be handled by Opus 4.8 instead of the new Mythos class system.

For crypto, the issue is indirect but relevant. Anthropic has no token, blockchain product, or DeFi partnership. But Claude models have become part of the tool stack for smart contract review, vulnerability research, and security analysis across digital assets.

If Fable 5 reroutes certain cybersecurity prompts to an older model, crypto security teams using Claude for smart contract auditing may not always get access to the frontier model’s strongest reasoning. That could matter in a sector where faster vulnerability detection can translate directly into prevented losses.

The launch also lands as Anthropic CEO Dario Amodei is calling for stricter AI oversight, including government power to block models deemed too risky to deploy. Fable 5 now gives Anthropic a live example of that philosophy in product form.