OpenAI May Add Watermarks to ChatGPT 4o Images; Details

OpenAI appears to be experimenting with the idea of watermarking images generated through its ChatGPT 4o image model — including those inspired by popular styles like Studio Ghibli. According to recent leaks, while the image generation feature will remain accessible to everyone, free-tier users may soon find watermarks on their downloads.

An AI researcher, @btibor91, shared insights on X (formerly Twitter), revealing that OpenAI's latest beta release for the ChatGPT Android app includes code referencing “image-gen-watermark-for-free,” suggesting watermarks may be added to images created by non-subscribers. This change was spotted in version 1.2025.091 2509108 beta and appears to be targeted at free users only.

The good news? Users subscribed to the ChatGPT Plus plan might still be able to save generated images without any watermark restrictions. However, OpenAI hasn’t officially confirmed the rollout of this feature, and there's a chance the company may revise or even drop the idea altogether.

In addition to the watermark development, OpenAI is reportedly working on introducing new tools. These could include features like “Structured Thoughts,” “Reasoning Recap,” “CoT Search Tool,” and something cryptically named “l1239dk1.” These tools are rumored to enhance the AI’s reasoning and search capabilities, though no specific launch dates or details have been officially announced.

Meanwhile, CEO Sam Altman shared further updates on OpenAI’s broader roadmap. After previously teasing the development of GPT-5, Altman has now mentioned that OpenAI is preparing to launch two intermediate models first — named o3 and o4-mini — in the coming weeks. He emphasized that these models would pave the way for a significantly improved GPT-5, stating, “We are going to be able to make GPT-5 much better than we originally thought.”

As for the ChatGPT 4o image generation tool, it continues to gain traction globally, with a particularly strong response from India. OpenAI’s COO Brad Lightcap recently called the launch “a very crazy first week,” revealing that over 700 million images have already been created since April 1. Of those, 130 million users engaged directly with the feature — a strong testament to its popularity.

India has quickly emerged as ChatGPT’s fastest-growing market. “The range of visual creativity has been extremely inspiring,” Lightcap shared. Sam Altman also praised the enthusiastic response from Indian users, calling it “amazing” and noting, “India is outpacing the world.” He even joined in on the fun by sharing an AI-generated anime-style image of himself dressed as a cricket player in a Team India jersey — a clever nod to the country's cricket-loving culture.