Microsoft - Policy Change
Executive Summary
Microsoft has restricted employee access to Anthropic's Claude Fable 5 AI model due to new data retention requirements that store user prompts and outputs for 30 days, with some flagged content retained for up to two years. The company's legal teams are evaluating whether the model can be approved for internal use amid concerns about protecting customer data and confidential information. Other Claude models remain available to Microsoft employees because they operate under zero data retention...
What Happened
On June 10, 2026, Microsoft restricted internal employee access to Anthropic's newly released Claude Fable 5 AI model due to data retention requirements that differ from other Claude models. While Microsoft deployed Claude Fable 5 to external customers through GitHub Copilot and Foundry, the company blocked it from the internal model picker that employees use. Anthropic's safety architecture for Fable 5 requires retaining user prompts and outputs for 30 days to operate safety classifiers, with content flagged as violating usage policies potentially stored for up to two years. Other Claude models remain available to Microsoft employees because they operate under zero data retention rules.
Who Is Affected
Microsoft employees are directly affected by losing access to Claude Fable 5 for internal work, while the company's GitHub Copilot and Foundry customers still have access to the model. Microsoft's legal teams are reviewing whether the data retention requirements create unacceptable risks for customer data and confidential company information that employees might input into the system. The restriction applies only to internal Microsoft use, not to the company's external AI service offerings.
Why It Matters
This incident illustrates how enterprise organizations must balance adopting cutting-edge AI capabilities against protecting sensitive business and customer information when AI providers retain user data. Microsoft's decision demonstrates that even partnerships between major technology companies face friction when data retention policies change, particularly when one party handles vast amounts of confidential information. The 30-day standard retention period and potential two-year retention for flagged content create meaningful compliance and confidentiality risks that major enterprises cannot ignore, even for advanced AI capabilities.
What You Should Do
If you use Claude Fable 5 through any platform, assume that Anthropic retains your prompts and the AI's responses for at least 30 days and avoid entering confidential business information, personal data, or proprietary content. If you work at an organization considering deploying Claude Fable 5, verify whether your legal and compliance teams have reviewed Anthropic's data retention policies against your data handling requirements. Check whether your AI service provider offers models with zero data retention options if your work involves sensitive information, and consult your IT or security team before using any AI tool with new data practices.
Summary generated from verified sources and reviewed before publication. How we summarize.