College students, beware: Anthropic will be adding watermarks to AI-generated text from its models, including the massively popular Claude.
Anthropic said that all models released after August 2 will automatically have tech to watermark both computer-generated text and files, as it announced on its website. Photo formats such as PNG and JPG will include metadata that shows that it was AI-generated.
"Because the watermark is part of the text, it will travel with the text when it's copied and pasted elsewhere, and may persist through some editing. Watermarking will be applied at the model level, which means it will be present no matter which Claude product or surface the text comes from," the support page update says.
Anthropic says a "supported" Claude model "weaves an imperceptible watermark directly into the text itself."
One of the most controversial aspects of the update is that the watermark will carry over to documents that were edited by Claude but were not originally AI-generated.
Per the website, "A detected mark provides a signal that content was processed by Claude, but is not fully conclusive. Detecting a Claude mark tells you that the content may have been processed by Claude. It does not, on its own, confirm the full provenance of the content."
"Claude may not be the original author. People often use Claude to proofread, translate, summarize, or convert files," the support page update notes.
Anthropic made the move to comply with the European Union's AI Act Transparency Code, which requires AI companies to mark AI-generated or edited content in a discernible manner. The tech company said it would release tools that would allow third parties to detect AI-generated content.
Some watermarks will not be detectable, including those in text pieces that are too short, are "heavily" edited, or were created prior to the new guidelines being implemented.
The move comes amid growing public scrutiny of AI as the technology spreads into daily life. Last week, the AI music platform Suno said it will flag music on its platform that is AI-generated.
Teachers have been coming up with new ways to detect AI usage, including one history professor who placed a hidden instruction in a question prompt, which generated nonsensical answers that made it obvious who was using AI to complete their exams. This trap led to 32 out of 35 of his students failing the midterm.
There's been ongoing discourse about the use of AI writing in the publishing industry, the media, the business world, and beyond. Last year, the Chicago Sun-Times was mocked for publishing an AI-generated list of recommended reading, which included several titles that didn't actually exist.
Create a free account to join the conversation!
Already have an account?
Log inSource link
