Anthropic Introduces EFS: Enterprise Safeguards
Anthropic launches Enterprise Frontier Safeguards (EFS) allowing customer-controlled cloud data storage alongside Claude Fable 5.1 and Mythos 5.1.

Stock photo for illustration only, not from the actual event
- Anthropic announces Enterprise Frontier Safeguards (EFS) for enterprise clients
- Enables customer custody, encryption keys, and human review in their own cloud
- Rolls out in phases with request-based access targeting broad availability this fall
- Introduces Claude Fable 5.1 and Mythos 5.1 reducing cache read costs by 75%
Anthropic has officially announced Enterprise Frontier Safeguards (EFS), a new architectural approach designed to balance data privacy and misuse detection for enterprise customers. Under EFS, monitoring data is stored in cloud infrastructure controlled directly by the customer rather than Anthropic servers. Detection remains handled by Anthropic, while custody, encryption keys, and human review stay entirely with the client.
The deployment of EFS is not available immediately. The architecture will roll out in phases with broad availability planned for later this fall through a request-based access model. Until official deployment, eligible customers can run Claude Fable 5.1 and Fable 5.1 under Zero-Data-Retention (ZDR) policies.
Anthropic noted that data retention is utilized for detection quality rather than model training. The company instituted a 30-day retention window starting with Fable 5 and reiterated that enterprise data has never been used for training without explicit authorization. Retaining data over a specific window is necessary because advanced security threats often span multiple tasks, sessions, and accounts, meaning instantaneous single-interaction classifiers cannot catch complex multi-step attack patterns.
The development of EFS involved collaboration with over 100 enterprise customers across financial services, healthcare, manufacturing, telecom, law, retail, and the public sector, alongside cloud providers AWS, Google Cloud, and Microsoft Azure. Contributors also included the Analysis and Resilience Center for Systemic Risk, featuring CISOs from major global financial institutions.

Stock photo for illustration only, not from the actual event
Anthropic's introduction of EFS marks a crucial milestone in overcoming security adoption barriers for highly regulated industries that typically face strict compliance hurdles preventing standard cloud AI adoption. By shifting data custody to customer-controlled environments while maintaining threat detection capabilities, enterprises can safely deploy advanced AI models without compromising internal security protocols.
"EFS is one of three enterprise concessions shipped alongside Claude Fable 5.1 and Mythos 5.1"
Anthropic
In addition to EFS, the release includes pricing and precision optimizations. Fable 5.1 reduced cache reads by 75% down to $0.25 per million tokens, cutting overall costs by roughly 25% on standard workloads and up to 45% on agentic tasks. Furthermore, updated cybersecurity safeguards generate about 60% fewer session interventions in Claude Code compared to the previous Fable 5 version.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment