Anthropic says it will watermark text generated by its AI models

Anthropic volition watermark substance generated by its models, including Claude, to comply with European regulations, the institution present says. The AI exemplary shaper confirmed the watermarking successful an updated enactment page.

EU AI Act’s Transparency Code, which took effect connected August 2, requires AI companies to people AI-generated oregon edited contented successful a mode different systems tin place them.

Anthropic said that each models released aft August 2 volition automatically person tech to watermark some computer-generated substance and files. For files, the institution is utilizing the C2PA unfastened standard.

The institution said it volition widen enactment for older models arsenic well, adding that the watermark volition question erstwhile users transcript and paste the text.

“Because the watermark is portion of the text, it volition question with the substance erstwhile it’s copied and pasted elsewhere, and whitethorn persist done immoderate editing. Watermarking volition beryllium applied astatine the exemplary level, which means it volition beryllium contiguous nary substance which Claude merchandise oregon aboveground the substance comes from,” the enactment leafage reads.

It’s not wide however overmuch editing users request to bash to region the watermark. We person asked Anthropic to clarify and volition update the communicative if we perceive back.

The institution noted that watermarking volition use to antithetic products similar Claude level API, Claude, Claude Code, Claude Cowork, and Claude Tag.

Platforms are present rushing to watermark AI-generated contented aft backlash from users and to debar regulatory scrutiny. Last week, AI euphony level Suno said it volition people tracks created connected its platform aft a spate of ineligible challenges. Last month, newsletter work Substack teamed up with Pangram to emblem AI-generated content. The company’s CEO, Chris Best, called retired Claudefishing, a word utilized for radical utilizing AI to make content.

Apart from Anthropic, different companies similar Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia person committed to adhering to the EU’s code.