# Provenance Marks on Model Output

Frontier vendors will keep extending default, invisible provenance marks across their model lines, and each rollout will produce removal tools and a dispute over who may run the detectors.

- Conviction: 40 / 100 (forming)
- Horizon: Emerging (watchlist)
- Tracking since: 2026-11-01T00:00:00.000Z
- Last updated: 2026-11-01T06:34:19.008Z
- Canonical: https://polylog.news/ai/trends/ai-output-provenance-marking
- Publisher: Polylog
- Affected regions: United States

## Recent evidence

- [confirms] Anthropic adds its hidden text watermark to Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6 on November 3 (2026-11-01): Anthropic will add its hidden text watermark to Mythos 5, Opus 4.7, Opus 4.6 and Sonnet 4.6 on November 3. On the two 4.6 models the mark applies only at default sampling settings, so changing one parameter removes it. This is the thesis in action: default marking spreads across a whole model line, and a built-in way to remove it ships at the same time.