Microsoft curbs internal use of Anthropic’s Claude Fable 5, citing data-retention requirements, report says

Microsoft has told employees to limit their use of Anthropic’s Claude Fable 5 over concerns about the AI startup’s data retention requirements, according to a report carried by Yahoo Finance and citing unnamed sources.

The restriction centers on how Anthropic will retain data submitted through the “Claude Fable 5” offering. In the report, Microsoft’s internal guidance is described as responsive to Anthropic’s new requirements, rather than a broader company-wide ban on using Anthropic models.

Anthropic, for its part, said in a statement reported earlier this week that it is rolling out the updated data retention policy. The timing matters because Microsoft’s guidance appears to follow after those changes were communicated to customers, prompting internal review of what data employees may be allowed to submit.

The report does not specify which Microsoft teams are affected, how the limits are enforced, or whether employees can access Claude Fable 5 via certain approved tools or only through specific, governed workflows. It also does not say whether Microsoft plans to replace the model with another option, or when any internal policy might change.

For Microsoft, the issue sits at the intersection of enterprise controls and AI product development. Large software companies run internal systems that often process sensitive customer and business information, so data retention rules become a practical question, not an abstract policy term.

Enterprise AI platforms typically need clarity on whether prompts and outputs are stored, for how long, and under what conditions they can be used for training or quality improvements. When providers introduce new retention terms, companies with strict governance requirements may reduce access until legal and security teams can evaluate residual risk.

The episode also reflects a broader market pattern. As more AI model vendors move from informal pilots to scaled deployments, contracts and technical controls increasingly determine which workloads are allowed, even within the same organization.

What Microsoft and Anthropic did not disclose in the reported account is as important as what they did. Neither the report nor any accompanying detail cited here provides figures on retention duration, whether retention differs by plan tier, or whether Microsoft negotiated carve-outs for specific internal use cases. Those specifics would largely determine how much practical impact the limits have across teams.