What happens to my data before Enterprise Frontier Safeguards ships?

Anthropic says eligible customers get zero data retention on Claude Fable 5 and Claude Fable 5.1 until Enterprise Frontier Safeguards is ready. Those organizations receive notice that they can use Fable under a zero data retention agreement in the meantime, so they are not blocked from the newest models while the phased rollout runs through the fall.

How much does Enterprise Frontier Safeguards cost?

Anthropic does not charge for Enterprise Frontier Safeguards. Because the detection data sits in the customer's own cloud account, the customer's cloud provider bills for the storage and data transfer it uses. There is no separate Anthropic line item, so the cost is whatever AWS, Azure or Google Cloud charges for holding that rolling window of traffic.

Which Claude products will support Enterprise Frontier Safeguards?

Anthropic lists Claude Code, Claude Enterprise, Claude Platform, Amazon Bedrock, Claude Platform on AWS, Google's Agent Platform and Microsoft Foundry as supported surfaces for Enterprise Frontier Safeguards. That spread means a team can keep the same data arrangement whether it calls Claude directly or through its existing cloud marketplace contract.

Will Anthropic staff read conversations that get flagged?

No. Anthropic describes Enterprise Frontier Safeguards as fully automated, with no Anthropic human review. Automated analysis runs over the customer-held data and sends serious misuse signals, such as offensive cyber or bio capability and stolen credentials, straight to the customer's own security team, which decides what to do next.

Why did Anthropic hold prompts for 30 days in the first place?

Anthropic's covered-model policy retains prompts and outputs for Mythos-class models for 30 days because some misuse only shows up across many requests. Coordinated jailbreaking or state-sponsored activity is invisible in a single call, so the detection needs a window it can correlate across sessions and accounts. Enterprise Frontier Safeguards keeps that window but moves it to the customer.

Anthropic · 2026-09-01 · major

Enterprise Frontier Safeguards — misuse checks run in your own cloud

Enterprise Frontier Safeguards runs Claude misuse detection on data kept in the customer's own AWS, Azure or Google Cloud account. The system replaces the 30-day retention rule for Mythos-class models, and Anthropic charges nothing for it.

Anthropic announcement illustration for Enterprise Frontier Safeguards

Anthropic moves Claude's misuse detection into the customer's own cloud account, so enterprises can run Fable and Mythos without 30-day retention.

Quick facts

Maker	Anthropic
Announced	1 September 2026
Replaces	30-day retention on Mythos-class models
Where data lives	Customer's own AWS, Azure or Google Cloud account
Human review	None by Anthropic; alerts go to the customer
Price	No Anthropic charge; cloud provider bills storage
Rollout	In phases, starting later this fall

What is it?

Enterprise Frontier Safeguards is a new arrangement that keeps Claude's misuse detection working while the underlying data stays in the customer's own cloud. Anthropic built it with more than 100 customers across financial services, healthcare, manufacturing, telecom, law, retail and the public sector, plus AWS, Google Cloud and Microsoft Azure. Named participants include Goldman Sachs, Morgan Stanley, Citi, Mastercard, Visa, Stripe, Snowflake and Salesforce.

How does it work?

Traffic is written to a bucket in the customer's own AWS, Azure or Google Cloud account, under keys the customer manages. Automated analysis then reads a rolling window of that traffic and looks for patterns that only appear across sessions and accounts. When it finds a serious signal, it alerts the customer directly rather than opening an Anthropic review queue.

Why does it matter?

Since June, Mythos-class models have carried a 30-day retention requirement on every platform that offers them, and that rule kept some regulated buyers off Claude Fable 5 entirely. Enterprise Frontier Safeguards keeps the safety window that the policy was protecting but puts it under the customer's own encryption keys, which is the version a bank or hospital compliance team can actually sign.

Who is it for?

regulated enterprise and public-sector teams

Frequently asked questions

What happens to my data before Enterprise Frontier Safeguards ships?: Anthropic says eligible customers get zero data retention on Claude Fable 5 and Claude Fable 5.1 until Enterprise Frontier Safeguards is ready. Those organizations receive notice that they can use Fable under a zero data retention agreement in the meantime, so they are not blocked from the newest models while the phased rollout runs through the fall.
How much does Enterprise Frontier Safeguards cost?: Anthropic does not charge for Enterprise Frontier Safeguards. Because the detection data sits in the customer's own cloud account, the customer's cloud provider bills for the storage and data transfer it uses. There is no separate Anthropic line item, so the cost is whatever AWS, Azure or Google Cloud charges for holding that rolling window of traffic.
Which Claude products will support Enterprise Frontier Safeguards?: Anthropic lists Claude Code, Claude Enterprise, Claude Platform, Amazon Bedrock, Claude Platform on AWS, Google's Agent Platform and Microsoft Foundry as supported surfaces for Enterprise Frontier Safeguards. That spread means a team can keep the same data arrangement whether it calls Claude directly or through its existing cloud marketplace contract.
Will Anthropic staff read conversations that get flagged?: No. Anthropic describes Enterprise Frontier Safeguards as fully automated, with no Anthropic human review. Automated analysis runs over the customer-held data and sends serious misuse signals, such as offensive cyber or bio capability and stolen credentials, straight to the customer's own security team, which decides what to do next.
Why did Anthropic hold prompts for 30 days in the first place?: Anthropic's covered-model policy retains prompts and outputs for Mythos-class models for 30 days because some misuse only shows up across many requests. Coordinated jailbreaking or state-sponsored activity is invisible in a single call, so the detection needs a window it can correlate across sessions and accounts. Enterprise Frontier Safeguards keeps that window but moves it to the customer.

Try it

Request access at https://claude.com/form/enterprise-frontier-safeguards