Microsoft Employees Face Limits on Claude Fable 5, Here’s What’s Behind It

Microsoft, despite being one of Anthropic’s biggest backers, is quietly restricting internal use of the company’s newest AI release- Claude Fable 5. The decision stems from Anthropic’s updated data retention requirements, which have raised red flags inside Microsoft over how employee and customer information might be handled.

Claude Fable 5, unveiled this week as Anthropic’s first “Mythos-class” model, was rolled out to GitHub Copilot and Foundry customers almost immediately. But internally, Microsoft employees noticed something unusual- the model was missing from the picker tool used for internal versions of GitHub Copilot. Other Claude models remain available because they operate under Anthropic’s Zero Data Retention (ZDR) rules, which guarantee that prompts and outputs are not stored. Fable 5, however, requires retention to power Anthropic’s new safety classifiers.

According to reporting from The Verge, Microsoft’s legal teams are currently reviewing Anthropic’s retention policies before clearing the model for internal use. The sticking point lies in how long Anthropic keeps data. Prompts and outputs are retained for 30 days by default, but if flagged for violating Anthropic’s usage policies, they can be stored for up to two years. For Microsoft, which handles sensitive customer data and proprietary code, that raises compliance and confidentiality concerns.

The company has reportedly told employees that its lawyers are evaluating whether these requirements align with Microsoft’s own standards for data protection. Until then, Claude Fable 5 remains off-limits internally, even as external customers gain access.

Anthropic’s Mythos-class models are designed to push the boundaries of AI capability. Just weeks ago, the company suggested the family was so advanced in cybersecurity tasks that it could be dangerous if released without safeguards. To mitigate risks, Anthropic introduced stricter prompt filters and safety classifiers but those safeguards depend on retaining user data, creating the current tension with Microsoft.

Microsoft declined to comment publicly on the matter. Still, the move highlights the delicate balance between innovation and trust in the AI industry. For employees, the restriction is a reminder that even cutting-edge tools can be slowed by legal and compliance hurdles. For Anthropic, it underscores the challenge of convincing partners that its safety-first approach doesn’t compromise confidentiality.

Claude Fable 5’s debut was meant to showcase the next leap in AI sophistication. Instead, it has sparked a debate inside one of its largest investors about where the line should be drawn between safety, privacy, and usability.

Read More: Anthropic's Claude Fable 5 Challenges GPT 5.5 and Gemini 3.1 Pro With Stronger Coding and Reasoning Skills