OpenAI introduces invisible text watermarks in European Union

OpenAI will begin adding an invisible watermark to text produced by ChatGPT and Codex for users in the European Union, according to TechCrunch and Digital Trends. TechCrunch and The Verge both report that the feature is rolling out over the coming weeks to eligible users across all ChatGPT and Codex plans in the region to comply with the European Union Artificial Intelligence Act. The European Union transparency rules required artificial intelligence companies to mark AI-generated content starting on August 2, TechCrunch reported. [1][6][7]

Developers using OpenAI's API globally can enable the watermarking feature for select models starting immediately, TechCrunch reported. The Verge notes that the invisible, machine-readable watermark is called textGrain. OpenAI stated that textGrain matched or exceeded rival methods like Google DeepMind's SynthID, according to The Verge. However, TechCrunch and Digital Trends both reported that editing text, short passages, or other modifications can make the invisible marks harder for systems to identify. [1][6][7]

In a separate development, OpenAI is expanding its advertising offerings by testing visual display ads in ChatGPT, GSM Arena, Engadget, The Verge, BleepingComputer, and Digiday all reported. The Verge, GSM Arena, and BleepingComputer report that the new ad format includes larger images and will put sponsored products and services on screen. According to The Verge and BleepingComputer, testing will begin in the United States later this month, initially appearing while users generate images. [2][3][4][5][10]

OpenAI originally introduced simple text-based ads between chat messages for free and Go users of ChatGPT in February, according to GSM Arena and The Verge. The company operates the chatbot service on a freemium model, offering text, audio, and image generation. ChatGPT originally launched in November 2022 and reached over 1.2 billion weekly active users by September 2026. [2][4][12][14]