Anthropic lets enterprises hold their own Claude data

Anthropic announced Enterprise Frontier Safeguards on September 1, a system built with more than 100 customers spanning banking, hospitals, factories, telecom carriers, law firms, retailers, and government agencies. It pairs two opt-in pieces: customer-owned storage, so Claude conversation data lives in the customer’s own Amazon S3, Azure Blob Storage, or Google Cloud Storage account under keys the customer manages rather than on Anthropic’s systems, and automated misuse monitoring that watches for signs of offensive cyber or biological-weapons development or stolen-credential use. When the monitoring flags something, the signal goes directly to the customer for review, with no Anthropic employee reading the flagged content first.

Anthropic built the system alongside security teams at Goldman Sachs, Morgan Stanley, Citi, Bank of America, Wells Fargo, Comcast, KPMG, Mastercard, Salesforce, and Visa, and is rolling it out in partnership with AWS, Google Cloud, and Microsoft Azure. There’s no separate charge for the safeguards themselves, customers pay their own cloud provider for storage, and the feature changes neither model behavior nor API pricing or rate limits. The phased rollout starts this fall; eligible customers get zero data retention on Claude models in the meantime while they wait for the full feature.

The design puts Anthropic in the position of courier rather than reader for this data, another instance of a pattern that’s shown up elsewhere this year: Anthropic ceding direct control over how its models get monitored and used in favor of giving that control to the customer or, in the Pentagon case, to the courts. It lands the same day as Anthropic’s Fable 5.1 and Mythos 5.1 model release, part of a busy September 1 for the company.