Anthropic Adds Watermark to AI Texts to Comply with EU Regulations

⚡

Key Takeaways

1Anthropic will integrate a watermark into the texts generated by its AI models, including Claude, to comply with European standards.

2The EU Transparency Code, in effect since August 2, requires the labeling of AI-generated content to facilitate its identification.

3Other companies like Google and Microsoft are also committing to follow these rules to avoid regulatory penalties.

💡Why it matters — This measure enhances the transparency of AI content, addressing legal requirements and user concerns.

Anthropic Adopts Watermarking for Its AI Models

The company Anthropic has decided to add a watermark to the texts generated by its artificial intelligence models, including Claude. This initiative aims to comply with new European regulations. The announcement was made via an update on the company's support page, confirming the integration of this technology.

Compliance with the EU Transparency Code

The Transparency Code, introduced by European AI legislation and in effect since August 2, requires companies to mark content generated or modified by AI. This marking should enable other systems to easily detect such content. Anthropic specified that all models released after this date will automatically incorporate this watermarking technology, applicable to both texts and files, thanks to the open standard C2PA.

Extending Watermarking to Older Models

Anthropic also plans to extend this feature to older models. The watermark is designed to move with the text when copied and pasted, and it may even persist despite certain modifications. According to the support page, watermarking will be applied directly at the model level, ensuring its presence regardless of the product or platform used with Claude.

Persistence and Limitations of the Watermark

Although the watermark is integrated into the text, the question of how many modifications are needed to remove it remains unclear. Anthropic has not yet clarified this point, but further explanations are expected. The watermarking will apply to various products, including the Claude platform API, Claude Code, Claude Cowork, and Claude Tag.

Reactions and Compliance from Platforms

In response to user criticism and to avoid regulatory scrutiny, many platforms are hastily adopting similar measures. Recently, Suno, an AI music platform, announced that it would mark its creations after facing legal challenges. Similarly, the newsletter service Substack has partnered with Pangram to signal AI-generated content. Substack's CEO, Chris Best, mentioned the term "Claudefishing" to describe the use of AI in content creation.

Commitment from Other Major Players

In addition to Anthropic, other tech giants like Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia have also expressed their commitment to comply with the EU code. This collective effort underscores the growing importance of transparency and compliance in the field of artificial intelligence.