AI-Generated Content Faces New Challenge with Claude Watermarking Rollout
A new tool from Anthropic is set to make it harder for people to pass off AI-generated content as their own. The company has announced that its Claude models will now embed an imperceptible watermark into text generated by the AI, making it easier to detect and investigate cases of plagiarism.
The watermark doesn’t change the meaning or readability of the original text but can be detected when copied and pasted. It may also persist through some editing processes, according to Anthropic. This move is part of the company’s efforts to increase transparency in line with the European Union AI Act.
A watermark is essentially a digital signature that proves ownership and authenticity of a piece of work. In this case, it will be embedded into text generated by Claude worldwide, including when accessed through cloud providers. Anthropic has also committed to providing third parties with tools to detect these watermarks.
The introduction of AI watermarking could have significant implications for the publishing industry, which has been grappling with concerns over AI-generated content. Last month, a book agent pulled support for a novel after allegations emerged that its author may have used AI to write it. The writer denied any involvement, but the incident highlights the need for greater transparency and accountability.
Anthropic’s move is not an isolated one - Google DeepMind has also been watermarking text generated by its Gemini app using SynthID technology since 2024. This development could give publishers, schools, and universities another tool to investigate cases of plagiarism and make it harder for people to pass off AI-generated writing as their own.
However, there are still potential loopholes in the system. Anthropic notes that heavy editing or paraphrasing can potentially remove the watermark, making it undetectable. Additionally, finding a watermark does not necessarily prove that Claude was the original author of the content, since even proofreading or translating with its AI can leave behind a mark.
The rollout marks a significant step towards greater transparency and accountability in the use of AI-generated content. As more companies begin to adopt similar measures, it will be interesting to see how this development impacts the publishing industry and beyond.