Summary - Microsoft has curtailed employee access to Anthropic’s Claude Fable 5 amid questions about the startup’s updated data retention rules. The move comes as Anthropic makes the most capable version of its Mythos-class model broadly available, touting strengths in software engineering and analytics, while placing explicit guardrails on uses considered risky, such as cybersecurity.
Microsoft has told employees to limit internal use of Claude Fable 5 while its legal teams evaluate Anthropic’s revised data retention requirements and the implications for customer and confidential information. The company has made no public determination yet on whether the model will be cleared for internal use.
Anthropic announced the public release of Claude Fable 5 as a version of its Mythos AI model family. The company described Claude Fable 5 as the most powerful model it has opened for wider use and highlighted its performance in areas such as software engineering and analytics. Anthropic also said it would enforce guardrails that prevent the model’s use in certain high-risk domains - cybersecurity being a cited example.
Under Anthropic’s stated retention policy for Mythos-class models, prompts submitted by users and the outputs generated are retained for 30 days across every platform where the models are offered, for trust and safety purposes. If those inputs or outputs are flagged by Anthropic’s trust and safety classifiers as violating the usage policy, the company said it retains that content for up to two years.
Microsoft’s internal concern focuses on the data retention provisions and how they intersect with customer data and confidential information. Legal teams are reviewing the changes; no decision has been announced on whether Claude Fable 5 will be permitted for internal workflows.
Anthropic also recently disclosed that it has confidentially filed for a U.S. initial public offering, without revealing the size or terms of the proposed offering. Both companies did not immediately respond to requests for comment.
Contextual note - The situation highlights the operational frictions that can arise when enterprise users and AI providers adopt differing policies on data handling and retention. For organizations that integrate external models into internal workflows, retention terms influence legal clearance and product adoption.