OpenAI will watermark ChatGPT outputs by default—but only in the EU

OpenAI has announced that from now on, text generated by ChatGPT will automatically carry a digital signature in the European Union. The move follows the EU AI Act, which has been in force since August and requires that content produced by artificial‑intelligence systems be identifiable by a method that other tools can detect. The company calls its technology “textGrain,” a proprietary process that embeds subtle patterns into word choices; these patterns are invisible to readers but can be picked up by a specialized detector that OpenAI will make available to a select group of researchers and organizations, with the possibility of expanding access through approval requests.

Outside the EU, the watermarking feature will be available but disabled by default, allowing users to choose whether to enable it. While standards such as SynthID and the C2PA project exist, the effectiveness of any watermarking method remains imperfect and can be circumvented by technically savvy users.

This initiative is significant because it aims to increase transparency and traceability of AI‑generated text, making it easier to spot automated content and helping to meet regulatory requirements. By offering the tool to researchers, OpenAI also encourages independent assessment of its reliability and potential adoption in other markets.

Read the original article on Ars Technica AI

This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.