Skip to content
EVENT / TOPICHistorical

Anthropic rolls out text watermark on older Claude models, but loophole exists on two variants

1 reports1 sources10/31/2026 updated

Current overview (not historical)

AI overview

Anthropic expanded text watermark support to four older Claude models (Mythos 5, Opus 4.7, Opus 4.6, Sonnet 4.6). However, the feature has a significant loophole: on Claude Opus 4.6 and Claude Sonnet 4.6, users can dodge the watermark by tweaking just one of three default sampling parameters. For Claude Mythos 5 and Claude Opus 4.7, the email announcing the rollout does not mention this sampling caveat, suggesting the watermark applies regardless of settings. This means the watermark is fully enforceable only when developers keep default parameters on the two affected models, while the other two models have no such escape hatch.

Written with AI from the reporting · updated 10/31/2026

Latest developmentNov 1, 03:20
Anthropic rolls out text watermark to four Claude models, but users can bypass it on two older versions
Explore the timeline, relations and evidence
Expand the detailed fact route and sources

Story route

A station-by-station route from the first public evidence onward; every station carries its own evidence and uncertainty.

  1. 1Unclassified fact stationNov 1, 00:00recorded Nov 1, 04:22single sourcerevised

    Anthropic watermark rollout leaves loophole on two older Cla

    Read original source ↗ The DecoderReplay this moment in the explorer
    Evidence & revisions · 1 source records
    Subject：
    anthropic
    Action：
    expanding text watermark to four more Claude models
    Object：
    Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6, and Claude Sonnet 4.6
    Occurred (fact side)：
    Nov 1, 00:00
    The Decoder ↗Report pageRevision 2 · currently revised
    Published (valid side)：
    Nov 1, 03:20
    First seen (known side)：
    Nov 1, 04:22
    Fact stationSingle source · one-sided claimRevised

    Copy may have been revised later; time replay is bounded. Layer labels group reading only, not causation between layers.

Uncertainty marks flag where to check the sources; they are not proven causation.

Report timeline

Explore every side through dated source reports.

Nov 1
  1. The DecoderRelated fact: single sourceSource report revised
    Anthropic rolls out text watermark to four Claude models, but users can bypass it on two older versions

    Report first recorded Nov 1, 04:22

    Anthropic is expanding its text watermark to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6, and Claude Sonnet 4.6. On the two older models, Claude Opus 4.6 and Claude Sonnet 4.6, the watermark only activates when developers leave three sampling parameters at default values; a minor tweak to any one of them allows users to bypass the mark. Anthropic cites the EU AI Act deadline as the reason for retrofitting older models, noting that every model released after August 2, 2026 ships with the watermark enabled by default.

    Read original source ↗

Event activity trend

Latest comparable
3
Comparable peak
10Nov 1, 05:00
24h change
-51%

The chart compares consistently observed participants only. Hover or tap to inspect each hour; use arrow keys to move between observations.

Event discussion

Reader opinions are hosted on GitHub Discussions; posting requires a GitHub account. Comments are not verified facts.

Open GitHub Discussions