Microsoft bans Claude Fable from employees over data protection issues

Anthropic released Claude Fableits first version of the Mythos-class AI, yesterday and it is already causing concern within Microsoft. Sources tell me that Microsoft is restricting the use of Claude Fable 5 to employees due to Anthropic’s new data retention requirements.

When Microsoft he quickly rolled it up Claude Fable 5 for his GitHub Copilot and Foundry customers, I am told that the model is not available in the model selector that Microsoft employees use in the internal versions of GitHub Copilot. All other versions of Claude are still available within Microsoft, as they operate under the Zero Data Retention (ZDR) rules.

I understand that Microsoft has been telling employees that their legal teams are reviewing changes to Anthropic’s data retention requirements. I’m told the biggest concerns are around customer data and privacy, and it’s unclear whether Microsoft’s legal teams will remove Claude Fable for internal use.

Claude Fable 5 it needs to save data using Anthropic’s new security measures, which means that Anthropic retains information and output and deletes this data after 30 days. Some information and output can be stored for up to two years if it is reported that it violates Anthropic’s policy, so there are legal concerns about Microsoft’s use of Claude Fable 5.

Claude Fable 5 is the first release from Anthropic’s Mythos class of AI models, and it arrives just weeks after the company announced that the family was capable of performing cybersecurity tasks so it did. too dangerous to be released publicly. Anthropic has implemented security measures to make Fable 5 more vulnerable, which made the change difficult for data storage.

We reached out to Microsoft for clarification on its use of Claude Fable 5, but the company did not respond in time for publication.