# Anthropic adds statistical text marks to two Claude models

_Tuesday, September 8, 2026 at 3:44 AM EDT · AI · Latest · Tier 2 — Notable_

Anthropic has begun embedding invisible statistical watermarks in output from its Fable 5.1 and Mythos 5.1 models, according to an account citing the company's August guidance.

The marks persist through copying and light editing, and a private-preview detection interface is limited to organizations with EU checking duties and certain listed groups. Anthropic says a detected mark indicates Claude participated in a passage, not that it authored it, and says worldwide use reflects its inability to region-limit the feature.

The company plans to add marking to older models over coming months.

## Sources

- [financialcontent](https://www.financialcontent.com/article/abnewswire-2026-9-8-claudes-ai-text-watermarks-raise-concerns-for-students-and-authors)

---
Canonical: https://techandbusiness.org/newswire/lfZaXxB7_ugazmVkz_04Lz
Retrieved: 2026-09-09T12:50:25.591Z
Publisher: Tech & Business (techandbusiness.org)