3 min readNew DelhiUpdated: May 24, 2026 03:49 PM IST
OpenAI is ramping up efforts to combat the spread of misinformation through AI-generated deepfakes with two new measures announced earlier this week.
The ChatGPT-maker on Tuesday, May 19, announced a new partnership with Google to include the invisible watermark called SynthID on all visuals generated using its AI products. Additionally, OpenAI launched a new public verification tool which allows users to easily find out whether an image was generated using AI or not.
The tool relies on two signals, namely SynthID and another open standard called C2PA, which adds a clear signal in the metadata of an image generated using AI.
To be sure, the tool is only designed to detect images generated with ChatGPT, the OpenAI API, or Codex. However, OpenAI has said that it is looking to expand it to cover other AI tools over time.
The move comes as AI image generators grow increasingly sophisticated and become widely available online. This, in turn, is making it harder to tell if an image is authentic. While the new protections announced by OpenAI could help combat the spread of misinformation, they may not be as impactful in stemming the flood of imagery coming from less reputable AI tools
“Watermarking can be more durable through transformations like screenshots, while metadata can provide more information than a watermark alone. Together, they make provenance more resilient than either layer would be on its own,” OpenAI said in a blog post.
To use OpenAI’s new AI image verification tool, follow these steps:
Story continues below this ad
-Upload a single image. Note, the formats supported by the tool are PNG, JPG, and WEBP.
-Review the results to see whether the tool detects C2PA metadata, a SynthID watermark, or no supported signal.
OpenAI further cropping screenshots closely around the image and avoid uploading files that contain multiple images to get the best results.
To test the tool, we uploaded an image generated using Images 2.0, OpenAI’s latest image generation model released in April this year. The response: “We did not find evidence that the content was generated using OpenAI tools. However, it may still have been AI-generated. See the FAQ for details.”
Story continues below this ad
Difference between C2PA and SynthID
The Coalition for Content Provenance and Authenticity (C2PA) is a non-profit that was founded in 2021 with a mission to mitigate the harmful effects of AI imagery on public discourse.
So far, the C2PA standard has been adopted by a range of Google products. However, adoption remains inconsistent across the industry. Additionally, the C2PA signal can be manipulated as it is easily accessible in the metadata of each file.
SynthID, on the other hand, was designed by Google DeepMind to persist even when bad actors attempt to remove it, either through screenshots, resizing or digital manipulation. While the two systems may individually have their own shortcomings, a tool that relies on both in a way that complements each other could prove to be more effective.
