From Anthropic · · 3m
Anthropic builds privacy safeguards with 100+ enterprise customers
Enterprise Frontier Safeguards lets companies store AI data in their own cloud while detecting misuse across sessions and accounts.
In brief
Enterprise Frontier Safeguards lets companies store AI data in their own cloud while detecting misuse across sessions and accounts. Anthropic's new Enterprise Frontier Safeguards lets companies keep AI conversation data in their own cloud infrastructure while automated systems monitor for misuse. Originally reported…
Privacy meets safety monitoring
Anthropic launches Enterprise Frontier Safeguards, combining zero data retention privacy with misuse detection. Data stays in customer-controlled cloud infrastructure.
Built with 100+ companies
Anthropic developed EFS with over 100 customers across financial services, healthcare, manufacturing, telecom, law, retail, and public sector.
Rolling out this fall
EFS launches in phases starting later this fall. Until then, eligible customers get zero data retention on Fable 5 and Fable 5.1.
More powerful, more risk
Mythos-class models like Claude Fable 5.1 bring major intelligence gains but also potential for misuse and autonomous misbehavior.
Sophisticated attacks need monitoring
Anthropic has seen attempted misuse ranging from fraud to cyberattacks. Some involve stolen credentials that spread across multiple sessions and accounts.
Why 30-day retention started
Detecting sophisticated misuse requires storing data to correlate patterns across time and accounts. That's why Anthropic introduced 30-day retention with Fable 5.
“This policy was not motivated by a desire to train on enterprise data: Anthropic has never trained on enterprise data without explicit permission, and never will.”
Regulated industries pushed back
Enterprises understood the security value of data retention but many in regulated industries couldn't use models that retained their data.
Banks helped design the solution
Anthropic worked with ARC, whose members include CISOs from Goldman Sachs, Morgan Stanley, Citi, Bank of America, and Wells Fargo.
Fortune 100 feedback shaped it
Anthropic consulted a quarter of the Fortune 100, every US global systemically important bank, and virtually every regulated industry.
Customers control the reviews
When automated monitoring detects suspicious patterns, signals go directly to customers. They review what the systems found, not Anthropic.
Data lives in customer clouds
Activity data used for monitoring can be stored in the customer's own cloud account, like Amazon S3, Azure Blob Storage, or Google Cloud Storage.
No Anthropic human review required
Automated systems analyze traffic for serious misuse signals, including cyber or biological threats and stolen credentials. Flags go straight to customers.
Wells Fargo keeps full custody
Wells Fargo says their logs stay in their environment under their keys. They keep data custody while Anthropic operates detection.
“Enterprise Frontier Safeguards gives us exactly what we asked for: our logs stay in a Wells-managed environment under Wells-managed keys.”
Banks set the requirements
Eight ARC members worked with Anthropic to define what it takes to run frontier models inside systemically important banks.
Opens up regulated workloads
The safeguards let companies apply AI in parts of the business they couldn't before, where work is regulated and highly sensitive.
Works across all platforms
EFS controls work the same whether you access Claude directly from Anthropic or through AWS, Google Cloud, or Microsoft Azure.
No extra charge from Anthropic
Anthropic doesn't charge for Enterprise Frontier Safeguards. Customers only pay their cloud provider for storage and data transfer.
The headline
Anthropic's new Enterprise Frontier Safeguards lets companies keep AI conversation data in their own cloud infrastructure while automated systems monitor for misuse.


