October 22, 2026, 12:40 pm | Read time: 3 minutes

A person in a Pink Panther costume flees from a water cannon, and in another image, two police officers carry him away. The images are said to have been taken during the demonstrations in France. Investigations by “BBC Verify” and the Spanish fact-checking team “Newtral” show, however, that the images were created or altered using artificial intelligence.

Here you will find content from third-party providers

To interact with or display third-party content, we need your consent.

Activate

In order to display embedded content, your consent to the processing of personal data, which can be revoked at any time (via the switch or via “Revoke tracking and cookies” at the bottom of the page), is required. Data may be transferred to third countries such as the USA (Art. 49 para. 1 lit. a GDPR). By switching to “on” you agree. You can find more information here.

Google’s SynthID Detector was among the tools used for verification. This tool detects invisible digital watermarks from certain AI systems. The trail for the water cannon image led to OpenAI, while the image with the two police officers led to Google. Apparently, real images from previous protests were partly used as a basis. Typical inconsistencies are also noticeable, such as illegible writing on a police vehicle and a crest that differs from the real badge of the French police.

Even an Alleged Assault Is Questionable

Another image shows an elderly couple in underwear on the street. The image is accompanied by the claim that protesters in Paris attacked them and stole their clothes. Whether such an incident actually occurred is currently unclear. The circulating image cannot substantiate it.

There are visible anomalies here as well. The thumb on the woman’s left hand is unusually large, and a masked figure on the right edge of the image has only one leg visible. VerificaRTVE further examined the image with SynthID Detector. According to this, it was created or altered using AI from Google. According to “Libération,” the Paris prosecutor’s office has not yet received any report of the alleged assault.

How to Check Suspicious AI Images

These examples show how difficult it can be to distinguish between real images and AI manipulations. One way to verify is through a reverse search with Google Lens. It can show where an image has previously appeared and in what context it was published. Reports from reliable media can also help to contextualize claimed events.

Also of interest: Germany Plans “AI Seepferdchen” for Elementary Students

Additionally, as described, SynthID Detector can be used to search for digital AI watermarks. However, this method also has limitations: If the tool does not find a corresponding mark, it does not automatically mean that an image is real. AI-generated or manipulated images can therefore remain undetected despite technical examination.