Anthropic scraps retention, offers customer-controlled EFS

Anthropic has reversed course on a controversial data retention policy following significant pushback from enterprise customers, particularly those in regulated industries. On September 1, 2026, the company announced Enterprise Frontier Safeguards (EFS), an offering that combines the privacy of zero data retention with automated safeguards for detecting misuse by storing activity data in cloud infrastructure controlled by the customer rather than by Anthropic. The system will be supported on Claude Code, Claude Enterprise, the Claude Platform, Amazon Bedrock, Claude Platform on AWS, Google’s Agent Platform, and Microsoft Foundry.

Security infographic showing automated threat detection and end-to-end encryption protecting data with customer-controlled...

What Changed and Why

Anthropic introduced its existing retention policy in June alongside the launch of Claude Fable 5 and Mythos 5, requiring 30-day retention of traffic on those models to help the company defend against misuse and what it described at the time as “complex and novel” cyberattacks. The policy met resistance from business customers concerned about how their sensitive information would be stored and reviewed.

Kate Jensen, Anthropic’s head of Americas, told CNBC that the company spent hundreds of hours working with customers to develop an alternative. Anthropic developed Enterprise Frontier Safeguards in collaboration with more than 100 customers across financial services, healthcare, manufacturing, telecom, law, retail, and the public sector, and worked with the Analysis and Resilience Center for Systemic Risk, whose members include the chief information security officers of Goldman Sachs, Morgan Stanley, Citi, Bank of America, and Wells Fargo.

Anthropic AI platform connecting Banking, Healthcare, Manufacturing, Legal, Telecom, Government, and Retail sectors with e...

How Enterprise Frontier Safeguards Works

The new system allows enterprise customers to store data in their own cloud infrastructure, including Amazon S3, Azure Blob Storage, or Google Cloud Storage, under their own encryption keys and access policies. Automated safety monitoring still scans for potential misuse, but no human review by Anthropic employees is required.

The approach represents a significant shift in how AI safety monitoring can be implemented. EFS stores monitoring data in cloud infrastructure the customer controls, not Anthropic’s, with detection staying with Anthropic while custody, keys, and human review stay with the customer. This architecture addresses the long-standing tension between enterprises needing zero data retention guarantees and security teams requiring misuse detection capabilities.

Anthropic Detection System architecture: AWS S3, Azure Blob, Google Cloud storage with encrypted data scanning and analysi...

Availability and Pricing

The system will roll out to customers in phases beginning later in the fall of 2026. Anthropic said it will not charge for Enterprise Frontier Safeguards. However, customers who choose their own cloud storage will need to cover their provider’s costs for storage, reads, writes, and data transfers.

Anthropic COBE FREE phased bollout timeline starting Fall 2022 with pricing comparison showing zero-cost EFS versus cloud ...

Customers eligible for the new system will receive zero data retention on Fable 5 and Fable 5.1 during the transition period. This interim measure provides immediate relief for enterprise customers awaiting the full EFS rollout.

Claude Code and New Model Releases

The Enterprise Frontier Safeguards announcement coincided with other significant product updates. Claude Code, Anthropic’s developer-focused offering, is among the platforms that will support the new safeguards system. Anthropic also unveiled its two latest models on Tuesday, Claude Fable 5.1 and Mythos 5.1, which the company said show improved coding, knowledge work and research capabilities.

Claude Fable 5.1 vs Mythos 5.1 comparison chart showing programming, problem-solving, and research capabilities.

Industry Context

Anthropic stated that it has never trained on enterprise data without explicit permission and never will. The company emphasized that the original retention policy was motivated solely by safety concerns, not by any desire to use customer data for model training purposes.

The move follows OpenAI’s introduction of its own safety monitoring system that does not require holding customer data. This indicates a broader industry shift toward privacy-preserving approaches to AI safety monitoring as enterprise adoption accelerates.

Jensen said she feels “very confident” that Anthropic will see accelerated deployment following the announcement, especially on customers’ most sensitive workloads. The change addresses a major barrier to adoption for enterprises in heavily regulated sectors that previously could not use frontier AI models due to data governance requirements.

Key Facts

Enterprise Frontier Safeguards announced September 1, 2026, replacing controversial 30-day data retention policy
Customers can store data in their own cloud infrastructure (Amazon S3, Azure Blob Storage, or Google Cloud Storage) with their own encryption keys
System developed with input from over 100 enterprise customers across regulated industries
No additional charge for Enterprise Frontier Safeguards; phased rollout begins later fall 2026
Supported platforms include Claude Code, Claude Enterprise, Claude Platform, Amazon Bedrock, Google’s Agent Platform, and Microsoft Foundry
Eligible customers receive zero data retention on Fable 5 and Fable 5.1 during transition period
New models Claude Fable 5.1 and Mythos 5.1 launched simultaneously with improved capabilities