Act on this, mostly facts

Future Claude models may embed text watermarks under EU AI Act

What to do

Check Anthropic's documentation on watermarking before relying on output provenance for future Claude models.

Why

An O'Reilly explainer cites an August 2026 announcement covering nine models, but the claim is secondhand.

What the research found

An O'Reilly explainer says Anthropic announced in August 2026 that its future Claude models embed text watermarks to comply with the EU AI Act, listing nine Claude models; the claim is secondhand and the only other hit found is a German report that OpenAI plans similar marking, so check Anthropic's own announcement before acting.

Claims and how well the evidence supports them

Anthropic announced in August 2026 that future Claude models will embed a text watermark in generated outputs 98%
The change is made to comply with the EU AI Act, which as of August 2 requires AI providers to mark AI-generated content 98%
Claude Desktop, Claude Code and the API outputs all contain the watermark 99%

Evidence

oreilly.com https://oreilly.com/radar/the-latest-anthropic-claude-models-contain-ai-text-watermarks-what-does-this-mean-for-you
news.ycombinator.com https://news.ycombinator.com/item?id=5189

Actionability rated medium.