Microsoft limits employee use of Claude Fable 5 over data storage rules

In short: Microsoft is restricting employees from using Anthropic’s Claude Fable 5 internally while it reviews Anthropic’s new rules for storing AI chat data.

What happened

Anthropic released Claude Fable 5, a new AI model, and it is already raising questions inside Microsoft. According to The Verge, Microsoft is limiting employee access to Claude Fable 5 because of changes in how Anthropic keeps data.

The model is not available in the internal “model picker” Microsoft employees use for internal versions of GitHub Copilot. GitHub Copilot is Microsoft’s AI helper for writing code (like an autocomplete tool that can also explain and draft code). Other Claude models remain available internally because they run under “Zero Data Retention” rules, meaning the AI provider does not store what employees type in or what the system replies.

At the same time, Microsoft has rolled out Claude Fable 5 to customers using GitHub Copilot and Microsoft Foundry. Foundry is Microsoft’s service for offering different AI models to businesses (like a menu of AI options). Microsoft told employees its legal teams are reviewing Anthropic’s updated data retention requirements.

Anthropic says Claude Fable 5 requires data retention so it can run new “safety classifiers” (automated checks that look for policy violations, like a spellcheck for harmful requests). Anthropic retains prompts and outputs and deletes them after 30 days. Some data may be stored for up to two years if flagged as violating usage policies.

Why it matters

This is a reminder that using AI at work can involve sharing sensitive information, even if it is unintentional. For businesses, rules about how long AI conversations are stored can affect privacy, customer trust, and legal risk.

Source: The Verge AI