Due to data retention policies, it’s been reported that Microsoft restricts employees from using Claude Fable 5 internally.

According to Beating Monitoring, due to concerns about Anthropic’s newly released model and its data retention risks, there are reports that Microsoft has restricted employees from using Claude Fable 5 in internal environments. Although Microsoft has rolled out Claude Fable 5 externally to GitHub Copilot and Foundry customers, the internal version selector used by employees does not offer Claude Fable 5.

Microsoft’s legal team is evaluating Anthropic’s data retention service terms. Unlike Claude models that Microsoft provides internally and that follow the Zero Data Retention (Zero Data Retention, ZDR) rule, Claude Fable 5 requires data retention in order to run a brand-new safety classifier. Anthropic will retain users’ prompts and output content for 30 days. If the content is determined to violate usage policies, the related data may be stored for up to 2 years, which could raise Microsoft’s legal concerns about leakage of customer data and confidential information.

As the first public release of Anthropic’s Mythos-level model, Claude Fable 5 was assessed before launch as being too powerful for network security tasks to be publicly released. To reduce security risks, Anthropic introduced a prompt safety protection mechanism; the operation of this protection mechanism directly brings about data retention requirements.