OpenAI has introduced an optional text watermarking feature through its API, a move aimed at helping the company comply with the European Union's AI Act. The tool lets developers mark AI-generated text, though it remains an opt-in capability. The company says the feature is designed to fit within the EU's transparency requirements, which are set to apply to general-purpose AI systems.
What the watermark API does
The feature is part of OpenAI's API offerings, not its consumer-facing ChatGPT interface. Developers can choose to activate it when generating text. The watermark itself is embedded in the output, though OpenAI hasn't detailed the technical method. The company describes the tool as optional, meaning developers aren't required to use it. That opt-in nature could limit how widely the watermark appears in real-world applications.
OpenAI's decision to offer the API comes as EU regulators prepare to enforce the AI Act's transparency obligations. Those rules require providers of general-purpose AI models to disclose how their systems work and to label AI-generated content. For OpenAI, adding a watermarking option is a direct way to show regulators it's taking those obligations seriously. The company hasn't said whether the watermark will become mandatory for any specific use cases.
Why the EU AI Act matters here
The EU AI Act is the first major law to set binding rules for AI systems across member states. It classifies AI applications by risk and imposes stricter requirements on general-purpose models like those OpenAI develops. Transparency is a key pillar: companies must make it clear when content is machine-generated. Watermarking is one method to achieve that. By offering an optional API, OpenAI gives developers a way to comply without forcing the feature on everyone.
But the word "optional" carries weight. If developers don't enable the watermark, AI-generated text could still circulate without any label. That gap is likely to draw scrutiny from EU authorities, who may push for stronger enforcement. OpenAI hasn't indicated whether it plans to make the feature default in the future. For now, the company is positioning the API as a tool, not a mandate.
A precedent beyond Europe
OpenAI's watermarking initiative could set a precedent for AI transparency standards globally. The EU's rules often influence tech regulations in other regions, and a major player like OpenAI adopting watermarking could encourage similar moves elsewhere. The company's approach—offering an optional API rather than a built-in requirement—might become a model for other AI developers. But it also raises questions about whether voluntary measures are enough to satisfy regulators in the US, UK, and Asia.
OpenAI hasn't announced partnerships with other companies or governments on this feature. The API is available now, according to the company, though it hasn't provided a specific rollout date for all users. Developers who want to use the watermark will need to check their API settings. There's no word yet on whether OpenAI will publish technical details or third-party audits of the watermark's reliability.
What developers and regulators watch next
The immediate test is whether developers actually adopt the optional watermark. If uptake is low, EU regulators may question the feature's effectiveness. OpenAI could face pressure to make it mandatory or to explain why it's optional. Meanwhile, the AI Act's enforcement timeline is ticking. Companies operating in the EU will need to demonstrate compliance, and OpenAI's API gives them one tool—but not a complete solution. The next few months will show whether voluntary watermarking becomes the norm or just a checkbox.
OpenAI hasn't said if it will extend watermarking to images, audio, or video. For now, the focus is on text. Developers building on OpenAI's platform should review their compliance obligations under the AI Act, especially if they serve EU users. The API is live, and the choice to use it is theirs. That choice may soon be judged by regulators.
