OpenAI Watermark Feature for AI-Generated Images

OpenAI is seemingly trialling a watermark feature intended for images produced by free users of ChatGPT, as noted in a recent beta version of their Android application. AI researcher Tibor Blaho discovered a reference marked “image-gen-watermark-for-free” in version 1.2025.091, indicating that watermarks may soon be specifically applied to images created on the free tier.

ChatGPT Updates: New Features and Watermarking

Recent updates to ChatGPT include a Student Plus referral programme now accessible to Colombian students from Universidad Nacional de Colombia, along with new mentions of “shared posts,” in addition to shared conversation, canvas, and deep research features in the web app. The new ImageGen watermark has also been noted in these updates.

Tibor Blaho (@btibor91) on Twitter

The image generation tool, introduced with the ChatGPT-4o update in early April, has swiftly become popular, especially for stylised images akin to Studio Ghibli art. However, concerns have arisen regarding the tool’s use for generating counterfeit documents, such as fake Aadhaar cards. The potential implementation of a watermarking system could serve as a measure to mitigate the dissemination of visual misinformation.

Current Status of the Watermark System

Presently, the watermarking system is in testing and has not yet been officially confirmed by OpenAI. Planned details may still be modified before any complete rollout, and it remains uncertain how noticeable or intrusive the watermark will be. Free users may continue generating such images, but they would receive downloadable versions marked with a visible watermark.

Impact on Free Users and Digital Content Sharing

If this feature is executed, it could significantly influence how free users distribute AI-generated images. This action appears to be a response to rising apprehensions regarding authenticity, copyright ambiguities, and the growing challenge of distinguishing real images from synthetic ones online.

Paid subscribers, like those enrolled in ChatGPT Plus, are unlikely to be impacted based on the current information leak, though OpenAI has not made any formal statements regarding this issue.