Anthropic on September 1, 2026, 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 roll out to customers in phases beginning later in the fall of 2026.
According to Anthropic, Mythos-class models such as Claude Fable 5.1 represent a major increase in intelligence and agentic capabilities, and that increase carries the potential for both misuse and autonomous misbehavior. The company said it has seen substantial evidence of attempted misuse of AI models, ranging from typical abuse such as fraud to sophisticated cyberattacks that can include agents autonomously engaging in destructive behavior. Some instances involve theft or misappropriation of enterprise customers’ credentials, which Anthropic said are difficult to detect without the ability to monitor traffic and spot abnormal behavior.
The most sophisticated misuse can involve many tasks spread across multiple sessions and accounts, Anthropic said, so running automated analysis on each interaction separately and then immediately discarding the data is not sufficient. Effective detection requires storing data for a meaningful period so it can be correlated across time and accounts. For that reason, the company introduced 30-day data retention starting with Fable 5. Anthropic said the policy was not motivated by a desire to train on enterprise data, and stated that it has never trained on enterprise data without explicit permission and never will.
The enterprises Anthropic consulted generally understood the safety and security value of data retention, the company said, but many, especially those in regulated industries, found it difficult to use models subject to data retention. Anthropic said it designed EFS with those customers to deliver the privacy of zero data retention alongside the detection benefits of monitoring across time and accounts.
Designed With More Than 100 Enterprise Customers
Anthropic said it developed EFS in close collaboration with more than 100 customers across financial services, healthcare, manufacturing, telecom, law, retail, and the public sector, and with its cloud partners Amazon Web Services, Google Cloud, and Microsoft Azure. The feedback came from the security, product, compliance, and delivery teams that will operate the system.
One group involved was the Analysis and Resilience Center for Systemic Risk (ARC), whose members include the chief information security officers of the largest US banks, among them Goldman Sachs, Morgan Stanley, Citi, Bank of America, and Wells Fargo. Anthropic said it also worked with leaders at companies including Comcast, KPMG, Mastercard, Salesforce, and Visa, and that the conversations spanned a quarter of the Fortune 100, every US global systemically important bank, and virtually every regulated industry.
“Enterprise Frontier Safeguards gives us exactly what we asked for: our logs stay in a Wells-managed environment under Wells-managed keys,” Wells Fargo said, adding that it keeps custody of its data while Anthropic operates the detection.
How the Controls Work
EFS is organized around three controls. On monitoring, when the system detects a pattern that needs attention, the signals are sent directly to the customer, whose own staff review what the automated systems detected. Anthropic said enterprises have long applied monitoring for insider risk and now want help extending that monitoring to agents, with concerns centered on whether automated monitoring systems meet their regulatory standards.
On storage, activity data used for monitoring can be kept in the customer’s own cloud account, such as Amazon S3, Azure Blob Storage, or Google Cloud Storage, under the customer’s own encryption keys, access policies, and audit logging. Anthropic said customers wanted their data to live in infrastructure they control rather than adding another trusted data vendor, which would require notifying their own customers and updating contracts.
On review, automated systems analyze a rolling window of traffic for signals of serious misuse, including attempts to develop offensive cyber or biological capabilities and signs of stolen or leaked credentials. Flags go directly to the customer, and no human review by Anthropic employees is required. Anthropic said many customers, particularly in regulated industries, operate under rules that tightly govern who may see certain information, such as privileged legal material, non-public information, and drug-safety reports, and that their teams are already trained and cleared for that review work.
Customer-owned storage, customer-managed encryption keys, and fully automated review are each opt-in, and none of them change model behavior, API pricing, or rate limits. The controls are designed to work the same way whether customers access Claude directly from Anthropic or through a cloud partner: customers on Amazon Web Services, Google Cloud, and Microsoft Azure receive equivalent controls, with activity data stored in their own cloud account. Anthropic said it is also working to support third-party offerings that serve EFS-eligible customers.
Pricing, Platforms, and Availability
Anthropic does not charge for EFS. Customers who elect to store their data in their own cloud account are billed by their cloud provider for that storage as well as for reads, writes, and data egress, the same way the provider bills any other resource.
EFS will be supported on Claude Code, Claude Enterprise, the Claude Platform, Amazon Bedrock, Claude Platform on AWS, Google’s Agent Platform, and Microsoft Foundry.
Anthropic said the phased rollout is aimed at making EFS broadly available later in the fall of 2026. Until EFS is ready, eligible customers will receive zero data retention on Fable 5 and Fable 5.1. Fable 5.1 is generally available across all three major cloud platforms and through the Claude API as claude-fable-5-1. Customers can request access to EFS through Anthropic’s request form.
Credit: Source link
