# Anthropic expands text watermarking to older Claude models

_Published Monday, November 2, 2026 at 4:18 PM EST · AI · Latest · Tier 2 — Notable_

![Anthropic expands text watermarking to older Claude models — Primary](https://s.yimg.com/lo/mysterio/api/489ec7a8af500567adf10c8e803be7dac3f761406f328407ec87e7eced35ab6b/lightyear_networkapi/resizefill_w1200%3Bquality_80%3Bformat_webp/https%3A%2F%2Fmedia.zenfs.com%2Fen%2Fthe_independent_577%2F00f0ed76e69ca55dbfbfd35cf75cb9d6.jpg)

Anthropic is extending text watermarking to older Claude models, including Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6, Yahoo Finance reported. The system embeds a pattern in generated words that automated detection tools can identify. The rollout covers users globally and has no opt-out control.

Detection is restricted to eligible organisations, including regulators, law enforcement, journalists and researchers, and companies required to verify watermarking under EU law. Short texts may contain too little of the pattern for detection. Anthropic says API settings that change word selection can remove the watermark on Opus 4.6 and Sonnet 4.6.

## Sources

- [Yahoo Finance](https://finance.yahoo.com/news/anthropic-claude-now-leave-trail-151343512.html)

---
Canonical: https://techandbusiness.org/newswire/G3NPS45N43GaJwDXINCU6G
Published: 2026-11-02T21:18:22.356Z
Story chronology: 2026-11-02T15:13:43.000Z
Retrieved: 2026-11-02T23:26:15.122Z
Publisher: Tech & Business (techandbusiness.org)