Claude will apply invisible watermarks to AI text and images

Authors and artists who use AI to generate text or images may find their work more easily traced. Anthropic, the company behind the Claude AI model, has announced that it will apply invisible watermarks to AI-generated content. This move comes as part of efforts to comply with the EU AI Act. According to The Independent, TechCrunch, The Verge, searchenginejournal.com and Yahoo Tech, the watermarks will help identify text and images created by AI, making it more difficult to pass off AI-generated work as human-created.

The watermarks are designed to be undetectable to the human eye but can be identified using specialized tools. This technology could significantly impact industries where originality and authenticity are paramount, such as publishing and digital art. The Verge and searchenginejournal.com note that the EU AI Act mandates such measures to ensure transparency and accountability in AI-generated content. The rollout of these watermarks raises questions about how effectively they can be implemented and detected.

The Independent and TechCrunch note that Anthropic's announcement does not specify the exact methods or technologies used for watermarking. Additionally, the potential for circumvention or evasion of these watermarks remains an open question.