Anthropic Unveils Enterprise 'EFS' to Simultaneously Implement Zero Data Retention and Abuse Detection

Anthropic announced Enterprise Frontier Safeguards (EFS). This solution combines the privacy protections of Zero Data Retention (ZDR) with the latest safeguards for abuse detection. Data is stored in cloud infrastructure controlled by customers, not Anthropic, and is scheduled for phased deployment starting this fall.

Resolving the Frontier Security Dilemma

Mythos-class models like Claude Fable 5.1 have significantly improved intelligence and agentic capabilities, but risks of abuse and autonomous malfunction have also increased. Attempts to misuse AI models, such as fraud, cyberattacks, and credential theft, have surged recently. In particular, complex abuse occurs across multiple sessions and accounts, making detection impossible with methods that immediately discard each interaction. Effective detection requires storing data for a certain period to analyze correlations across time and accounts.

Customer-Centric Design

Anthropic developed EFS in collaboration with over 100 customers in regulated industries such as finance, healthcare, and manufacturing, as well as AWS, Google Cloud, and Microsoft Azure. The existing 30-day data retention policy was necessary for safety but became a barrier to adoption for customers in regulated industries. EFS was designed to reflect these needs, allowing customers to directly control where and how their data is stored and managed.

• Monitoring Control: When automated systems detect anomalous patterns, they send signals directly to customers, allowing them to conduct their own reviews.
• Data Storage Control: Customers can store and manage data in their existing cloud infrastructure, reducing the burden of registering with additional trusted data vendors.

EFS is supported on major platforms including Claude Code, Claude Enterprise, Amazon Bedrock, Google’s Agent Platform, and Microsoft Foundry.