OpenAI will begin including an invisible watermark to textual content generated by ChatGPT and Codex within the European Union to adjust to the EU AI Act, the corporate mentioned Monday in a blog post.
The EU AI Act’s transparency rules, which took impact on August 2, require AI firms to mark AI-generated content material in a manner different techniques can establish.
OpenAI mentioned the watermark will roll out over the approaching weeks to eligible ChatGPT and Codex customers on all plans, however solely within the EU. Builders utilizing OpenAI’s API anyplace on this planet can flip it on for choose fashions beginning at the moment; it’s off by default. OpenAI mentioned it’s not making textual content watermarking a worldwide default at launch.
The watermark is just not an precise image, however works by subtly shaping the mannequin’s phrase selections, leaving a sample readers can’t see, however a detector can choose up. As a result of it lives within the phrases themselves, it travels with the textual content when it’s copied and pasted. OpenAI mentioned the watermark doesn’t establish the person, and that it noticed no significant change in its fashions’ efficiency with it switched on.
OpenAI additionally revealed a technical report for its technique, referred to as textGrain, alongside the announcement. Co-written with researchers from the College of Pennsylvania and Yale, it walks by an instance of utilizing a secret key to type next-word predictions to complete the sentence. Add a whole bunch of those nudges collectively, and the detector can spot AI-generated content material utilizing solely the textual content and the important thing.
Can the watermark be eliminated by enhancing? OpenAI’s assessments counsel sure. In a single check, changing 10% of phrases with synonyms dropped detection from about 92% to 66%. The corporate additionally mentioned brief passages, math solutions, and translated textual content are tougher to detect.
“These limitations contribute to our resolution to offer preliminary detector entry solely to accepted researchers and professional organizations, who can assist us consider reliability and accountable makes use of,” mentioned the corporate.
OpenAI additionally cautioned {that a} lacking watermark “doesn’t show human authorship.” The textual content might be too brief or too closely edited, or it may come from one other firm’s AI.
“[Watermarks] can point out that an OpenAI system generated or processed a part of a passage, however not how a lot human judgment, enhancing, or creativity went into it,” the corporate mentioned.
The announcement comes two months after Anthropic said it would watermark textual content generated by Claude, a transfer it’s making use of worldwide. That call drew backlash from some Claude users, who argued they’d provided “the directions, context, selections” whereas Claude was simply “the device.”
OpenAI had constructed a textual content watermark earlier than however held off on releasing it, partly over issues that customers would swap to rivals that didn’t watermark, The Wall Street Journal reported in 2024.
Anthropic, Google, Meta, Microsoft, and OpenAI are among the many firms which have dedicated to following the EU’s code of practice on AI-generated content material.
Once you buy by hyperlinks in our articles, we may earn a small commission. This doesn’t have an effect on our editorial independence.
