Anthropic Plans Data-Retention Shift to Ease Enterprise Concerns

Anthropic’s 30-day data retention policy for Claude enterprise users had raised concerns about how sensitive corporate information could be stored, accessed and reviewed. The issue is particularly important for customers in regulated industries, where privacy controls and audit requirements can determine whether generative AI tools are approved for broader deployment.

Anthropic has introduced Enterprise Frontier Safeguards, giving corporate customers greater control over data storage and review procedures. The system supports automated safety monitoring without requiring human reviewers to inspect customer content. The policy revision follows customer pushback over the 30-day retention framework and is intended to reduce privacy and security barriers to enterprise adoption of Claude.

Anthropic previously offered zero data retention, or ZDR, arrangements to some enterprise and cloud-service customers, under which prompts and model outputs were not stored long term. It is changing that approach for Claude Mythos 5 and Fable 5 to analyze anomalous patterns across multiple interactions and guard against jailbreak attacks and large-scale abuse. The shift has implications for enterprise data governance and privacy compliance.

Anthropic has announced that prompts and outputs from Claude Mythos 5 and Fable 5 will be retained for 30 days, primarily affecting enterprise and cloud-service customers previously covered by ZDR. Microsoft has temporarily restricted employees' internal use of the advanced models over data-privacy concerns. The available information does not disclose when the policy will take effect, how many customers will be affected or any related financial amounts.