Skip to content

Security and governanceBrief

Claude's text watermark: what caching and editing change

Anthropic says future Claude models generate text that contains a statistical watermark. A gateway that caches, trims, translates or reroutes that text decides how much of the signal survives. This brief sorts the common operations and gives a reviewer a short, accurate statement.

By INFRO Editorial

Published

4 min read

A grid of dark glass cards with small bar charts, a few of them lit amber, on a dark ground

In short

The watermark lives in word choices, so verbatim copies keep it, while truncation, light edits and non-Claude fallbacks dilute or remove it.1 Treat it as probabilistic evidence of Claude involvement, never as proof of authorship, and record which model and which edits touched each stored text.

Key takeaways

  • The watermark is a pattern in which words Claude picks, not added characters, so an exact copy of a response keeps it.
  • Truncation, light edits and proofreading shrink the evidence; a full rewrite removes it, and detection is weak on short samples.
  • A translation by Claude is watermarked, but a summary or rewrite by a different model carries that model's marking, not Claude's.
  • Log the serving model and every post-processing step next to stored text, so a later detection result can be read correctly.
  • A watermark says Claude was probably involved; it says nothing about ownership, authorship or who sent the request.

What the watermark is, and what it is not

Anthropic says future Claude models will generate text that contains a watermark, a way to estimate the likelihood that Claude was involved in writing it. It says it is doing this to comply with the EU AI Act.1 The marking obligations entered into application on August 2, 2026, and about 190 organisations, Anthropic among them, signed the Code of Practice on Transparency of AI-generated Content.2

The mechanism matters for a gateway. Nothing is added to the text and there are no hidden characters. The method is a version of SynthID-Text: it changes the source of the randomness used to pick among near-equivalent words, using a key and a few preceding words.1 So the signal is a statistical pattern across many word choices. It survives whatever preserves those words, and weakens as they are removed or replaced.

Limits Anthropic states: detection works poorly on small samples, confidence grows with length, factual passages and code carry less watermark, and the check only answers "how likely was Claude involved". It cannot say a text is human-written or that another AI wrote it.1

Which gateway operations keep, dilute or drop it

The "Effect" column follows from Anthropic's description of the mechanism. The reading for stored copies is our inference, not a measured result.

OperationEffect on the watermarkNote for reviewers
Exact-match cache replayKept: the stored text is the original answerSame text served again, possibly to many users
Truncation or length cutoffDiluted: fewer word choices left to testShort fragments may not be detectable
Light edit or proofreadingMostly kept in long text, weak in shortDetection can be too weak to register1
Full rewrite of every wordRemoved1Arguably no longer AI-generated text
Translation by ClaudeKept: every word is Claude's choice1Translating with another tool replaces it
Summary by ClaudePresent, since Claude chooses the wordsShort summaries give little to detect
Fallback to a non-Claude modelNot Claude's watermark; that model has its own key or method1Record which model answered
Code outputLess watermark; comments may carry some1Do not rely on it for code provenance

Two further cases. Concatenating Claude text with human or other-model text yields a mixed document, and the watermark can only attest to the Claude-chosen portion. Redaction by a guardrail changes words in place; the result is edited Claude text, not the original.

Caching, logging and fallback in practice

An exact-match cache returns the stored answer verbatim, so the watermark is preserved on every hit. INFRO's response cache is opt-in and exact-match, and it only stores a response when output guardrails left it untouched and every choice finished with stop or tool_calls.3 That means cached entries are original provider text, not edited text, and a truncated answer (finish reason length) is never replayed.3 If you build your own cache, decide explicitly whether it may store post-processed text, and label which kind each entry is.

Logging can truncate. When INFRO records a request body, content_truncated is true if the stored copy was shortened, and a long string keeps its beginning and end around a marker naming how many characters were cut.4 If you plan to run detection on archived answers, that logged copy is not the original text. Payload logging is off by default and retention is yours to choose, so confirm what you actually keep before promising anything.

Fallback changes the model. Connection failover is same-model, but model fallbacks are cross-model and opt-in, and the response names the model, provider and connection that served the request.5 Store that field beside the text. A reviewer who later gets a negative Claude detection result needs to know the answer came from a different model, not from an unmarked Claude.

What to tell a reviewer

  • The gateway does not add, remove or verify watermarks. It passes through whatever the provider returns, plus any edits you configure.
  • Verbatim copies (cache hits) retain the pattern; truncated, redacted, merged or rewritten copies may not.
  • A detection result is probabilistic. A positive means Claude was likely involved at some point; it cannot separate writing from heavy editing, and a negative does not prove human authorship.1
  • Detection access is limited. Anthropic describes a private-preview detection API for eligible organizations and for enterprises obliged to verify watermarking for their own compliance; others can register interest.1
  • Ownership and legal responsibility do not change because of a watermark.1

Open questions to put to Anthropic or your cloud provider: whether older Claude models you still call are marked yet (Anthropic describes a transition period for models launched before August 2, 2026), and whether Claude served through a cloud platform carries the same marking. Anthropic's page does not answer the second.1

Questions readers ask

Does the watermark change token counts or price?
Anthropic says watermarking produces no extra tokens and the model costs the same to serve and use, so estimated cost per request should not move because of it.1
Can a watermark be traced to our organization or a user?
Anthropic states the watermark and its key carry nothing that would let anyone recover information about the user, their organization or their chats.1

From the team that builds it

Where this shows up in INFRO

INFRO is a control plane in front of the provider accounts you already hold. The capabilities below are the parts of it this article touches, each with the documentation that defines its behaviour.

  • Response cache

    Available with a plan

    Opt-in, exact-match cache for chat completions. An identical request inside the time-to-live you set is answered from the cache and never reaches your provider, so your provider does not bill it. Every plan; the longest time-to-live depends on the plan. No semantic matching.

    Read the documentation for Response cache
  • Request traces and payload logs

    Available with a plan

    Per-request timing detail, with prompt and completion logging off by default and a retention window you choose.

    Read the documentation for Request traces and payload logs
  • Automatic failover

    Available with a plan

    A route that times out or 5xxs before the response starts is retried elsewhere without your client seeing it.

    Read the documentation for Automatic failover
  • Guardrails

    Available with a plan

    Pattern-based checks on chat completion prompts and responses: block, redact or flag email addresses, phone numbers, card numbers, IBANs, US Social Security numbers, IP addresses, API keys and your own terms, plus optional classifier checks on your own connection to a decision model's provider. Detection can miss things. Included on every plan.

    Read the documentation for Guardrails

INFRO accounts are self-serve. Provider usage is billed to you by your providers; INFRO charges a software subscription and takes 0% of model spend. See plans.

Sources

Numbered in order of first citation. Each was opened and checked against the claim on the date shown.

  1. 1.How Claude's text watermark works (opens in a new tab) — Anthropic. Accessed Nov 8, 2026.
  2. 2.Strong backing for the Code of Practice on Transparency of AI-generated Content (opens in a new tab) — European Commission. Accessed Nov 8, 2026.
  3. 3.Response cache (opens in a new tab) — INFRO. Accessed Nov 8, 2026.
  4. 4.Request traces (opens in a new tab) — INFRO. Accessed Nov 8, 2026.
  5. 5.Failover & fallbacks (opens in a new tab) — INFRO. Accessed Nov 8, 2026.

About this article

Read Anthropic's watermark explainer (published Aug 14, 2026, updated Sep 1), the European Commission's code-of-practice news item, and INFRO's response cache, traces and fallbacks docs on Nov 8, 2026. Anthropic's page does not describe cloud-platform behavior or the final status of every Claude model, so those are listed as questions to ask, not stated as fact.

Last checked against its sources on . The next review is due by . No separate named reviewer is recorded for this article.

How articles are written, sourced and corrected is set out in the editorial standards.