OpenAI adding invisible text watermarks to ChatGPT in the EU

By Splitfeed EditorialLast updated Oct 13, 2026, 14:40 UTC

OpenAI announced in mid-October 2026 that it will apply invisible watermarks to text generated by ChatGPT for users in the European Union. The rollout was disclosed alongside more than 20 other new features announced in the same release window. The watermarking applies specifically to EU users in connection with compliance requirements under the EU AI Act 2026, and reports confirmed that users in Malaysia will not receive the feature enabled by default.

The EU AI Act sets out disclosure requirements for AI-generated content, and this watermarking feature represents one of the first concrete technical steps a major AI provider has taken to meet those obligations at scale. Invisible text watermarking, which embeds signals into generated output without visible markers, is considered a more practical compliance tool than visible labels, but its reliability and resistance to removal remain open questions in the field. The geographic split, EU users receiving it by default while Malaysian users do not, illustrates how regulatory pressure rather than product philosophy is driving the rollout.

None of the three camps have published reactions to this story yet. Pro-AI voices would typically frame this as responsible corporate behavior, evidence that the industry can self-regulate when given clear legal targets, and a feature that protects users without hampering capability. Anti-AI voices would typically argue that invisible watermarks are insufficient, that they are easy to strip or defeat, and that the EU-only scope shows companies respond to disclosure requirements only when legally compelled. Middle Ground observers would typically call the move a meaningful but incomplete step, raising questions about enforcement, auditability, and whether watermarks alone satisfy the spirit of the AI Act.

The practical test will come as the EU AI Act 2026 requirements move toward enforcement. How regulators assess whether invisible watermarking meets the disclosure standard, and whether independent researchers find the watermarks robust or easily removed, will determine whether this announcement is treated as genuine compliance or a minimal gesture.