Anthropic Adds Watermarks to AI-Generated Text to Comply with EU Regulations
Anthropic, the company behind the popular AI model Claude, has announced that it will be watermarking text generated by its models. This move is in response to the European Union’s (EU) new regulations on artificial intelligence, which require companies to clearly identify AI-generated or edited content.
The EU’s Transparency Code, which took effect on August 2, mandates that AI companies mark their generated or edited content in a way that can be easily identified by other systems. To comply with this regulation, Anthropic will be applying watermarks to all text generated by its models, including Claude.
All models released after August 2 will automatically include the watermarking technology, which will identify both computer-generated text and files as AI-created. For files, the company is using the C2PA open standard, a widely accepted method for verifying digital content’s authenticity.
The watermark will not only be present in new models but also extended to older ones, ensuring that all users are compliant with the EU regulations. Additionally, when users copy or paste text from an AI-generated source, the watermark will travel with it, persisting through some editing processes.
According to Anthropic’s support page, ‘the watermark is part of the text, so it will travel with the text when it’s copied and pasted elsewhere.’ However, it remains unclear how much editing users need to do to remove the watermark. We have reached out to Anthropic for clarification on this matter and will update our story accordingly.
Anthropic’s decision to implement watermarks is not unique; several other companies are also taking steps to comply with the EU regulations. Last week, AI music platform Suno announced that it would be marking tracks created on its platform after facing backlash from users and regulatory scrutiny. Newsletter service Substack has also partnered with Pangram to flag AI-generated content.
Other notable companies like Black Forest Labs, Google, Meta, Microsoft, OpenAI, and Synthesia have committed to adhering to the EU’s code, demonstrating a growing trend towards transparency in the AI industry.