CONFIRMED

Anthropic announced Enterprise Frontier Safeguards (EFS) on September 2, 2026, a feature that combines the privacy of zero data retention with safeguards for detecting misuse of Claude in corporate accounts. The announcement was published on Anthropic's official blog and confirmed by CSO Online, Help Net Security and MarkTechPost.

How it works: the data stays with the customer, not with Anthropic

The core point of EFS is where monitoring data is stored. Instead of living on Anthropic's servers, activity data used to detect misuse now lives in cloud infrastructure controlled by the customer, whether Amazon S3, Azure Blob Storage or Google Cloud Storage, under the encryption keys, access policies and audit logging the customer already uses. Anthropic describes the design as provider-side zero data retention combined with customer-custodied retention.

What the system detects, and who sees the alerts

Automated systems analyze a rolling window of traffic for signals of serious misuse, including attempts to develop offensive cyber or biological capability and signs of stolen or leaked credentials. When such a signal is identified, the alert goes directly to the customer and does not pass through human review by Anthropic employees, leaving it to the customer's own team to decide what to do next.

Why Anthropic is doing this now

According to the company itself, the decision comes from substantial evidence of attempted misuse, ranging from ordinary fraud to sophisticated cyberattacks, including cases of theft or misuse of enterprise customers' credentials. EFS tries to resolve a real tension: corporate customers want maximum privacy over what they send to the model, but Anthropic also needs some mechanism to flag serious abuse of its own platform. The proposed solution relocates where the evidence is kept, rather than eliminating the need to keep evidence.

When it arrives and for whom

The feature will roll out in phases across this Northern Hemisphere fall, between September and December 2026, offered to eligible customers, with no single confirmed general availability date so far.

Why it matters for Brazilian agencies and SMBs

Agencies that process end customer data through Claude, for customer service, conversation analysis or marketing automation, often have to choose between zero retention, which protects privacy but makes internal auditing harder, or standard retention, which helps investigate incidents but keeps sensitive data on a third party's server. A model where monitoring logs stay in the company's own custody, rather than Anthropic's, could ease LGPD compliance while preserving the ability to investigate internal misuse. It is worth tracking EFS's rollout this fall and checking with Anthropic's sales team whether the feature will already be available for smaller or mid-sized accounts before the end of 2026.

Sources

Anthropic, Developing Enterprise Frontier Safeguards with our customers: https://www.anthropic.com/news/enterprise-frontier-safeguards | CSO Online, Anthropic introduces zero-retention AI safety monitoring for enterprises: https://www.csoonline.com/article/4217538/anthropic-introduces-zero-retention-ai-safety-monitoring-for-enterprises.html | Help Net Security, Anthropic's Enterprise Frontier Safeguards lets your Claude logs stay in your cloud: https://www.helpnetsecurity.com/2026/09/02/anthropic-enterprise-frontier-safeguards/