EU AI Act watermarking obligations begin for new models
New model outputs carry an invisible watermark, with a detection API for vetted groups.
Under the EU AI Act’s Code of Practice on transparency, models released after 2 August 2026 must watermark their outputs. Anthropic signed alongside around 190 other organisations. The watermark is invisible without a detection API, which is being offered to vetted groups.
It is a small but concrete governance step: an obligation that is in force, not merely agreed.
Risk 0 · Mitigation 2.