Anthropic will embed invisible watermarks in all text generated by Claude models launched after August 2, 2026. This measure aims to comply with Article 50 of the EU AI Act, which requires AI providers to mark their generated content so it can be identified. The watermarks, invisible to the naked eye, will be embedded in the statistical patterns of the texts and should not affect their quality. This requirement will apply to all Claude platforms worldwide, including the API, chat interface, and Claude Code. Anthropic has also committed to providing transitional marking support for older models launched before the August 2026 deadline.
Source: Read the original article
