Microsoft Restricts Employee Access to Anthropic’s Fable 5 AI Over Data Retention Concerns

Microsoft has blocked internal employee access to Anthropic’s Claude Fable 5 AI, citing concerns over its updated data retention policies. The Mythos-class model, which is publicly available to customers through GitHub Copilot and Microsoft Foundry, requires prompts and outputs to be stored for 30 days to support safety monitoring, though flagged data may be retained for up to two years. The decision comes despite Microsoft’s broader push to integrate AI tools across its workplace ecosystem. Employees currently lack access to Fable 5 through Microsoft’s internal GitHub Copilot platform, unlike other Claude models that operate under zero-data-retention policies. This restriction follows Microsoft’s earlier move to discontinue Claude Code subscriptions for employees, though financial considerations are not the main driver in this case. Anthropic’s retention policy—necessary for identifying misuse like jailbreak attempts—has raised compliance and legal concerns for enterprise customers handling sensitive data. Microsoft’s legal and compliance teams are reviewing the framework, leaving uncertain whether Fable 5 will later be approved for wider internal use. Other Claude models remain accessible internally due to their zero-retention arrangements. The move has sparked mixed reactions online, with some joking about productivity impacts, while others question Microsoft’s motives, suggesting cost may still play a role. The situation highlights the tension between adopting advanced AI and meeting strict enterprise compliance standards as AI adoption grows. Microsoft’s cautious approach underscores the challenges large organizations face when balancing access to cutting-edge AI tools with protecting proprietary and customer data. The company’s review process will determine whether Fable 5 gains broader internal approval or remains restricted.