---
title: "OpenAI Adds Optional Text Watermarking to API to Meet EU AI Act Requirements"
description: "OpenAI is introducing optional text watermarking for its API to comply with the European Union's AI Act, which mandates machine-readable provenance for AI-generated content."
author: "CryptoResearch AI"
published: "2026-10-05T16:03:50.411Z"
updated: "2026-10-05T16:03:50.412Z"
category: "ai"
reading_time_minutes: 1
content_type: "editorial"
canonical: "https://cryptoresearch.news/news/openai-adds-optional-text-watermarking-to-api-to-meet-eu-ai-act-requir"
tags: ["OpenAI", "EU AI Act", "Artificial Intelligence", "Regulation", "Technology"]
---

# OpenAI Adds Optional Text Watermarking to API to Meet EU AI Act Requirements

> Editorial content, written by CryptoResearch.

OpenAI is introducing optional text watermarking for its API to comply with the European Union's AI Act, which mandates machine-readable provenance for AI-generated content.

OpenAI is rolling out new text watermarking features for users in the European Union to align with the bloc's AI Act. The update, which includes an optional text watermark API, is designed to meet transparency requirements under Article 50 of the legislation.

The EU AI Act, which became law in 2024, requires generative AI outputs to carry machine-readable markings. While the rules took effect on August 2, 2026, companies have until December 2, 2026, to ensure their systems are compliant.

OpenAI’s approach to the mandate involves a multi-layered strategy. For images and audio, the company is using C2PA metadata and SynthID watermarks, the latter of which is licensed from Google DeepMind. For text, developers can now enable watermarking features within their API project and organization settings.

The company has opted to make text watermarking an optional feature rather than a default setting. This decision follows years of internal hesitation regarding the technology. In 2024, OpenAI noted that while an internal prototype was 99.9% effective, the company had concerns regarding the robustness of text watermarks and the potential for unfair flagging of non-native English speakers who use AI tools for writing assistance.

Other industry players have moved more quickly to implement similar measures. Google has offered SynthID-Text in its Gemini models since 2024, while Anthropic began using imperceptible text watermarks in its new Claude models on August 2, 2026.

The shift places the responsibility on developers and businesses building on OpenAI’s platform to integrate these tools into their own products. As the December deadline approaches, the industry is waiting for further clarification from regulators on whether opt-in tools will be sufficient to meet the full scope of the EU's provenance requirements.

## Sources

- [Crypto Briefing](https://cryptobriefing.com/openai-text-watermark-eu-chatgpt-ai-act/)

---

Published by CryptoResearch. Canonical version: https://cryptoresearch.news/news/openai-adds-optional-text-watermarking-to-api-to-meet-eu-ai-act-requir