OpenAI introduces EU text watermarking and tests new ChatGPT visual ads

OpenAI announced on Monday that it will start adding invisible watermarks to text generated by ChatGPT and Codex in the European Union. TechCrunch and Digital Trends reported that the update is designed to comply with the transparency requirements of the EU AI Act, which took effect on August 2. TechCrunch added that the watermark will roll out over coming weeks to eligible users on all plans within the EU. [1][5]

The Verge reported that the textGrain watermarking system matched or exceeded competing approaches such as Google DeepMind's SynthID. However, TechCrunch and Digital Trends both noted that OpenAI acknowledged the invisible marks can become harder to detect if the text is edited or altered. For API developers globally, TechCrunch reported that the watermarking feature can be enabled starting today for select models, though it remains off by default. [1][5][6]

Alongside regulatory compliance updates, OpenAI is expanding its commercial ad offerings. The Verge, qz.com, and GSMArena.com reported that OpenAI is testing visual display ads in ChatGPT. According to The Verge, the latest format will display images of sponsored products and services on screen, with testing set to begin in the United States later this month. [2][4][9]

Digiday reported that the new visual ad formats arrive alongside updated measurement tools and brand suitability testing as OpenAI seeks to attract larger advertiser budgets. According to The Verge, these visual ads will initially appear when users generate images using ChatGPT, building upon the ad formats the company first introduced to the platform in February. [4][7]