Anthropic released Claude Fable 5, its first Mythos-class model, and Microsoft has immediately restricted internal access to the system. While the technology is available for external customers on GitHub Copilot and Microsoft Foundry, employees cannot select it from the internal model picker. This exclusion stems from new data retention requirements introduced by Anthropic for the latest release. The company mandates that certain user data be retained for analysis and model improvement, a policy that directly conflicts with Microsoft’s internal Zero Data Retention rules. Consequently, all previous Claude models remain accessible to staff because they operate under the older, stricter privacy framework. The new restriction highlights a growing friction between enterprise security protocols and the evolving operational needs of leading AI providers.
This situation matters because it signals a potential shift in how large technology firms manage third-party AI integration. Microsoft’s decision to prioritise internal data sovereignty over accessing the latest generative capabilities suggests that privacy compliance will increasingly dictate technology adoption. If other organisations follow this lead, the pace of innovation for enterprise AI could slow as vendors struggle to reconcile performance with data governance. The incident also underscores the risks of relying on external models when their underlying policies change without prior notice. For security teams, it serves as a reminder that even trusted partnerships require rigorous vetting of data handling terms before deployment.
- Microsoft blocked internal access to Claude Fable 5 due to a conflict with its Zero Data Retention policy.
- The restriction applies only to employees, while external customers retain full access to the new model.
- Enterprise adoption of third-party AI may face new hurdles as data retention rules become more stringent.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
