Microsoft restricts Claude Fable for employees over data retention concerns
Microsoft is limiting employee access to Anthropic’s Claude Fable 5 while its legal teams review new data retention rules.
Intelligence analysis by GPT-5.4 Mini
Microsoft has shipped Anthropic’s Claude Fable 5 to external customers, but is holding it back from internal employee use because Anthropic now retains prompts and outputs under its safety setup. The company’s legal teams are reviewing whether that retention policy is compatible with Microsoft’s handling of customer and confidential data.
Microsoft got a new AI tool, but it’s holding it back for its own workers because the tool keeps copies of chats for a while. It’s like a school letting a new locker system in for visitors, but checking the rules carefully before using it for staff files.
Analysis
What happened
The Verge reports that Microsoft is limiting employee access to Anthropic’s Claude Fable 5 because of Anthropic’s new data retention requirements. Microsoft has already made the model available to GitHub Copilot and Foundry customers, but it is not showing up in the model picker used by Microsoft employees for internal versions of GitHub Copilot.
Why Microsoft is cautious
According to the report, Microsoft’s other Claude models remain available internally because they operate under Zero Data Retention rules. Claude Fable 5 is different: Anthropic’s new safety classifiers require retained data, so prompts and outputs are stored and then deleted after 30 days. The article says some prompts and outputs can be kept for up to two years if they are flagged as violating Anthropic’s usage policy.
That creates a legal review problem for Microsoft, which is worried about customer data and confidential information. The company has reportedly told employees that its legal teams are evaluating Anthropic’s updated retention requirements, and it is not yet clear whether internal approval will come through.
Broader context
The report also notes that Claude Fable 5 is Anthropic’s first broad release from its Mythos class of models, arriving only weeks after Anthropic said the family was so capable at cybersecurity tasks that it was too dangerous to release publicly. Anthropic says it has added prompt safeguards, but those safeguards are tied to the new retention rules that are now creating friction for Microsoft.
Microsoft declined to comment.
Key points
- Microsoft is limiting employee use of Claude Fable 5 over Anthropic’s data retention requirements.
- The model is still available to GitHub Copilot and Foundry customers.
- Microsoft’s other Claude models remain available internally under Zero Data Retention rules.
- Anthropic retains prompts and outputs for 30 days, and some flagged data can be stored for up to two years.
- Microsoft’s legal teams are reviewing whether the model can be used with customer and confidential data.
If Microsoft’s legal review clears the model, Claude Fable 5 could still become available for internal use after the retention concerns are resolved. That would let Microsoft keep offering the model to customers while adding it back into employee tools.
If Microsoft decides the retention policy is too risky, Claude Fable 5 may stay out of internal Copilot workflows for some time. That would limit adoption inside Microsoft and keep the company focused on models that can operate under zero data retention rules.
