OpenAI announced a phased rollout of text watermarking for its ChatGPT and Codex models in response to the EU AI Act, marking a significant step toward reliable digital provenance. The new invisible watermark will help identify AI‑generated text, giving journalists, researchers and the public a machine‑readable signal of a piece’s origin. API customers worldwide can now opt in to this feature, while the watermark is being automatically added to eligible outputs in the European Union over the coming weeks. OpenAI is also opening a limited‑access program for researchers to use its watermark detector, aiming to refine the technology through real‑world feedback. These moves aim to make the creation and editing history of digital content transparent, addressing growing concerns about authenticity and the spread of misinformation. [[1]]