Microsoft stopped letting employees use Anthropic's Claude Fable 5 in internal Copilot environments.
The model, released yesterday as Anthropic's first Mythos-class AI, was quickly rolled out to GitHub Copilot and Foundry customers. Internally, however, the model no longer appears in Microsoft’s model picker. The company says the change is due to Anthropic’s new data retention requirements, which conflict with Microsoft’s Zero Data Retention (ZDR) policy. Other Claude models remain available because they comply with ZDR.
The move highlights the growing friction between AI providers’ data policies and enterprise security rules. Companies that rely on external AI services may now have to audit retention clauses before granting internal access, limiting the speed of AI adoption.
So far, the restriction is limited to staff; customers can still use Claude Fable 5 through Microsoft’s external offerings.