Korea AI News

OpenAI Introduces Invisible Watermarks in ChatGPT and Codex Text in the EU

AI NewsWave | Oct 6, 2026, 16:55 KST | Based on 4 Korean news reports

Key points

• OpenAI adds invisible watermarks to generated text
• Measure responds to the EU AI Act
• Applied to the EU first, not worldwide at once

OpenAI is beginning to insert invisible watermarks into text generated by ChatGPT and Codex. The measure, disclosed on the 5th (local time), makes no difference to a person reading the text but lets machines confirm that it is AI-generated.

The watermark will be applied to all ChatGPT plans and more in the EU within the next few weeks. AI Times reported that the measure responds to the European Union's AI Act and is not being applied worldwide at once.

AI Times reported that OpenAI is adopting a similar approach after Anthropic.

Sources (in Korean)

• 테크M
• AI타임스
• 아이뉴스24
• 디지털투데이

Compiled by AI NewsWave from facts confirmed across the Korean reports above. Translated from the Korean version.