AnswersPublished22 sources
Anthropic EFS lets enterprises keep Claude activity logs in their own cloud
Anthropic’s Enterprise Frontier Safeguards (EFS) combines zero data retention style privacy with automated misuse detection: enterprise activity data stays in the customer’s cloud, under its keys and policies, rather... EFS is designed for regulated organizations that need cross session safety monitoring but cannot...
Published byEdited with GPT-5.6 LunaImages generated with GPT Image 1.5
Research answer
Key takeaways
- Anthropic’s Enterprise Frontier Safeguards (EFS) combines zero data retention style privacy with automated misuse detection: enterprise activity data stays in the customer’s cloud, under its keys and policies, rather...
- EFS is designed for regulated organizations that need cross session safety monitoring but cannot allow a model provider to retain or manually review sensitive prompts and outputs.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What is Anthropic’s Enterprise Frontier Safeguards framework, why did the company introduce it after its June 30-day retention policy trigge. Article summary: Enterprise Frontier Safeguards (EFS) is Anthropic’s enterprise architecture for combining zero-data-retention-like privacy with automated detection of serious model misuse. Rather than retain monitoring data itself, Anth. Topic tags: general, general web, user generated. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fa
Anthropic’s Enterprise Frontier Safeguards (EFS) is a customer-controlled architecture for monitoring misuse of its most capable Claude models without requiring Anthropic to hold enterprise activity logs. The company says EFS combines the privacy of zero data retention (ZDR) with automated safeguards that can detect patterns across related interactions. 6
The key change is where monitoring data lives: instead of being retained on Anthropic’s systems, it can remain in storage controlled by the enterprise, with the customer managing the encryption keys, access policies and audit records.
Why Anthropic introduced EFS
Anthropic introduced a 30-day retention requirement for traffic to covered frontier models beginning June 9, 2026. The stated reason was safety monitoring, including the need to investigate complex and novel misuse that may only become visible across multiple sessions or accounts. 1
23
That policy created a problem for regulated businesses. Organizations such as banks and other companies handling sensitive information often have strict requirements governing where data is stored, who can access it and how it is reviewed. Vendor-held retention could make those models difficult or impossible to approve for sensitive workloads. 1
17
EFS is Anthropic’s proposed compromise: preserve automated, cross-session misuse detection while giving the customer custody of the monitoring data and responsibility for any human investigation.
How Anthropic EFS works
Under EFS, activity data used for safety monitoring can be stored in the customer’s own cloud account, including:
- Amazon S3
- Azure Blob Storage
- Google Cloud Storage
The data is governed by customer-controlled encryption keys, access policies and audit logging. Anthropic’s automated systems analyze a rolling window of traffic for signals associated with potentially serious misuse, such as attempts to develop offensive cyber or biological capabilities or indications of stolen credentials. 4
5
When the system identifies a concern, the resulting flag is sent to the customer. Anthropic says the workflow can operate without Anthropic employees reviewing the underlying customer content; the customer’s own team handles any human review or follow-up. 1
4
Anthropic says EFS will be available without an additional charge. 1
15
EFS was developed with regulated customers
Anthropic says it worked with more than 100 customers across regulated and other industries while developing EFS. The company says those discussions included roughly one-quarter of the Fortune 100, every U.S. global systemically important bank and eight members of the Analysis and Resilience Center for Systemic Risk. 6
The named participants include Comcast, KPMG, Mastercard, Salesforce and Visa. 6
Wells Fargo’s CISO, Munish Kumar Sharma, said the framework addressed the bank’s request for logs to remain in a Wells-managed environment under Wells-managed keys. In that arrangement, Wells Fargo retains custody while Anthropic provides the detection capability—an approach Sharma said could allow the bank to use frontier models while meeting its responsibilities to customers, employees and regulators. 6
Supported Claude and cloud platforms
Anthropic says EFS will support:
- Claude Code
- Claude Enterprise
- The Claude Platform
- Amazon Bedrock
- Claude Platform on AWS
- Google’s Agent Platform
- Microsoft Foundry
6
This matters because the controls are intended to work whether an organization accesses Anthropic’s models directly or through supported cloud and platform partners. 1
6
When will EFS be available?
EFS is not broadly available yet. Anthropic says it will roll out the framework to customers in phases beginning later in fall 2026, with broad availability targeted for the fall. 3
6
The timing creates an interim arrangement for eligible enterprise customers. Until EFS is ready, those customers can receive zero data retention for Claude Fable 5 and Fable 5.1 under Anthropic’s ZDR terms. Access is not automatic: customers must apply and be approved. 6
7
For covered models without an applicable ZDR or EFS arrangement, Anthropic’s policy retains prompts and outputs for 30 days to support safety work. 1
23
How EFS compares with OpenAI’s Private Safety Processing
Anthropic’s EFS and OpenAI’s Private Safety Processing address a similar enterprise problem: how to detect misuse patterns across related interactions without giving provider employees routine access to the underlying prompts and responses. 30
31
The clearest EFS distinction is its explicit customer-cloud custody model. EFS is described as keeping monitoring data in the enterprise’s own storage account, under the enterprise’s keys and access controls, while Anthropic’s systems perform automated detection. 4
6
OpenAI’s published design also supports customer-controlled infrastructure for ZDR deployments. OpenAI says it is separately developing an option in which customer content would be stored on OpenAI infrastructure but encrypted with customer-controlled keys. In both configurations, its automated systems are intended to return limited safety signals without exposing the underlying content to OpenAI personnel. 30
These are company descriptions of evolving architectures, not a substitute for reviewing each provider’s final technical, contractual and compliance terms.
What the policy means for consumer plans
EFS is an enterprise framework rather than a consumer privacy setting. Anthropic’s support documentation says the covered-model retention update does not affect Claude Free, Pro or Max plans across Claude’s web, desktop and mobile apps, including Claude.ai and Claude Code. 23
For enterprise buyers, the practical distinction is different: eligible organizations can apply for interim ZDR on Fable 5 and Fable 5.1, then move to EFS when it becomes available. Other covered-model deployments remain subject to the applicable 30-day safety-retention terms. 6
8
23
The bottom line
EFS shifts the enterprise AI privacy trade-off from “retain data with the provider or lose advanced safety monitoring” toward a customer-controlled model. Anthropic’s automated systems can still look for coordinated misuse, but the enterprise controls storage, keys, access and human review. The important caveat is timing: EFS is being rolled out in phases, and interim ZDR requires customer approval rather than applying automatically. 1
6
7
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Anthropic EFS lets enterprises keep Claude activity logs in their own cloud"?
Anthropic’s Enterprise Frontier Safeguards (EFS) combines zero data retention style privacy with automated misuse detection: enterprise activity data stays in the customer’s cloud, under its keys and policies, rather...
What are the key points to validate first?
Anthropic’s Enterprise Frontier Safeguards (EFS) combines zero data retention style privacy with automated misuse detection: enterprise activity data stays in the customer’s cloud, under its keys and policies, rather... EFS is designed for regulated organizations that need cross session safety monitoring but cannot allow a model provider to retain or manually review sensitive prompts and outputs.
Sources
cnbc.comAnthropic changes data retention policy after pushback ...
support.claude.comData retention practices for Covered Models - Claude Support
cxtoday.comAnthropic Drops Data Retention Requirement for Mythos and Fable 5.1 After Enterprise Pushback