Event Published Models and productsAnthropic
Anthropic will keep enterprise Claude logs in the customer's own cloud, at no charge
Anthropic will hold the enterprise Claude traffic it screens for misuse in the customer's own cloud under the customer's keys, at no charge, and says prices and rate limits do not move.
What happened
Published here
On 1 September 2026 Anthropic announced Enterprise Frontier Safeguards, an arrangement under which the traffic used to spot misuse of Claude is held in the customer's own cloud account, under the customer's keys, instead of on Anthropic's systems. Automated systems read a rolling window of that traffic for signals such as stolen credentials, and anything they flag goes to the customer, with no review by an Anthropic employee. Anthropic says it designed the arrangement with more than 100 customers, charges nothing for it, and that it changes neither model behaviour, nor API prices, nor rate limits. It reaches customers in phases from later this autumn, and until then eligible accounts keep zero data retention on Fable 5 and Fable 5.1. (Anthropic announcement, 1 September 2026)
Read the full description of this diagram
Diagram in three parts. Two panels sit either side of a dashed custody boundary: on the left, a box for the rolling window of enterprise Claude traffic held under Anthropic's own keys, and on the right the same rolling window in the customer's own cloud account under the customer's own keys, outlined as the thing that moved. Below each window an identical box marked unchanged reads that window for signals such as stolen credentials, and the right-hand panel ends with flags going straight to the customer with no Anthropic employee in the path. A band across the bottom sets Anthropic's charge of zero against no change to model behaviour, API prices and rate limits, and against the customer's new cloud bill to store, read and move those logs plus the staff hours of whoever answers the alerts.
Tracked on this site: our Claude API price table
What they are saying
Outlets read
CNBCa company backing away from its own rule
Anthropic on Tuesday announced it will change a controversial data retention policy for its business customers after "a lot of feedback," and it said it plans to implement a new solution called Enterprise Frontier Safeguards instead.
read
Help Net Securityreading it as the security team who has to run it
Anthropic does not say how long that window runs, which matters if you are trying to reason about what an attacker could hide by spacing activity out.
read
Our verdict
The disagreement is about what kind of event this is. CNBC hears a vendor giving way to the people paying it, while Help Net Security takes the design at face value and asks what it leaves unsaid, which is how far back the detection can look. Both readings can be right at once, and the number that would settle the second question has not been published.
What it means for you
This is a retention change dressed as a product, and the reason it exists is that the retention it replaces was blocking sales. Read against our own table, nothing moves: Claude Fable 5.1 is still recorded at $10 per million input tokens and $50 per million output, with cache reads at $0.25, as read on 5 September 2026. Anthropic says plainly that these controls leave prices and rate limits where they are. So the cost of the change is not on the invoice you get from Anthropic. It is on the one you get from your cloud provider, which now bills you for storing, reading and moving the logs, and on the staff hours of whoever answers the alerts that used to be somebody else's job.
In more detail
The numbers, and where each one comes from
- More than 100 customers worked on the design, across financial services, healthcare, manufacturing, telecom, law, retail and the public sector, per Anthropic’s announcement.
- 30 days: the retention window Anthropic introduced with Fable 5 and that this arrangement is built to replace, from the same page.
- $0 charged by Anthropic for the arrangement itself, with storage, reads, writes and egress billed by the customer’s cloud provider instead.
- $10 and $50 per million input and output tokens for Claude Fable 5.1, with cache reads at $0.25, as our table recorded them on 5 September 2026, unchanged by this announcement.
Who this is actually for
Read the eligibility line rather than the headline. This is enterprise work: named cloud partners, customer-managed keys, an alert queue somebody has to staff. A person paying twenty dollars a month for Claude is not in it, and Anthropic’s post does not claim otherwise.
The interesting move is where the cost went. Anthropic charges nothing, which reads well, and the storage bill lands with Amazon, Google or Microsoft, who charge for holding the logs and for taking them out again. That is a real cost with a real invoice; it is just not Anthropic’s invoice, and a buyer comparing vendors on price alone will not see it.
What is still not published
The monitoring reads a rolling window. Nothing published says how long that window is, and the length is the whole strength of the mechanism: correlation across sessions only works over the period you can still see. Our Claude page will keep the price rows dated as they are; if a window length is ever published, it belongs beside them rather than in a blog post nobody can date.
Sources
Sources read
Anthropic announcement, 1 September 2026 (anthropic.com)
Newsroom read
our Claude API price table, verified 5 September 2026, which this change does not move (aijourney.space)
Own ledger read
Cite this
AI Journey, "Anthropic will keep enterprise Claude logs in the customer's own cloud, at no charge", 6 Sep 2026, https://aijourney.space/updates/2026-09-06-anthropic-enterprise-frontier-safeguards/ (source: Anthropic, verified 6 Sep 2026)
Next in the stream
Anthropic cancels the Claude Sonnet 5 price rise that was due on 1 September
Anthropic withdrew a scheduled September price increase for Claude Sonnet 5, leaving its introductory $2 input and $10 output rates as the model's standard API prices.
Previous
Nvidia agrees to buy Hugging Face, and writes open model rules into its risk factors
Nvidia has agreed to buy Hugging Face for about $11.9 billion, and its SEC filing lists government limits on distributing open models as a risk to the platform.
Next
More in Anthropic
Study Finds Precedent Labels Sway Claude Haiku Approval Calls
A new paper tests Claude Haiku 5.5 alongside other models on refund and claim decisions, showing that past case labels shift approvals sharply.
Paper Tests Claude Opus 4.8 Agents on Conflicting Code Edits
A new paper benchmarks Claude Code running Claude Opus 4.8 against other coding agents on tasks where two agents edit the same shared codebase at once.
Paper Finds Claude Opus 5.5 Agent Beats Automated CT Labeling Harness
A new paper testing CT scan labeling finds a free-form agent running Claude Opus 5.5 outscored an automated review harness on most of six public datasets.