Thai police station distributes its own AI photo of arrest by officers in glitter dresses

A photo of five male police officers and a woman in colourful, glittering dresses surrounding an arrested suspect drew international media attention in early June 2026. The image turned out to have been created using artificial intelligence and distributed by the police station itself. Sergeant Rachata Mitrsuripong, who manages the Facebook page of the Tha Luang Provincial Police Station in Thailand's Lopburi province, admitted this to news agency AFP on 26 May.

The arrest behind the photo was genuine. Officers from the station detained a drug dealer suspected of drug trafficking and illegal gambling activities. During the arrest they seized 53 methamphetamine pills, more than 200 small plastic bags and a mobile phone. The original photo showed five male officers in ordinary clothing surrounding the suspect. That more straightforward version never appeared on the Facebook page.

Instead, on 21 May Sergeant Rachata posted an AI-generated version he had produced himself, depicting the officers in festive glitter dresses. Several major international outlets, including The Telegraph, the New York Post, the Daily Mail and CNN Brazil, initially reproduced the image as authentic.

A friendlier image as the goal

Sergeant Rachata explained his motivation to AFP. According to him, the purpose of the edited image was to project a friendlier image of the police, to show a playful and humorous side so that citizens would feel more comfortable approaching officers. The station's Facebook page serves as a communication channel to the local community, and a lighthearted tone apparently fits that strategy.

The fact that the image was not labelled as AI-generated meant that international media reproduced it uncritically as coverage of a real operation. It was only after questions arose about the implausible clothing that the sergeant explained what had happened. The station itself did not publicly initiate that clarification; the admission came in response to questions from AFP.

How the photo misled international media

The spread of the image illustrates how quickly AI-generated visuals can enter mainstream news streams via social media and press reports. Newsrooms at outlets including The Telegraph, the Sun, the Mirror and the Daily Star published the photo without verifying whether it was authentic. CNN Brazil and the New York Post did the same.

Visually, there were indications that, on closer inspection, pointed to a generated image, such as the unusual clothing in a police operation context, but none of the newsrooms involved established this before publication. After AFP brought the facts to light, corrections and amendments followed at several outlets.

The incident fits into a broader pattern of cases in which AI-generated images, including those originating from official or semi-official accounts, are shared without clear labelling and subsequently picked up as news. The AI tool Sergeant Rachata used to generate the image has not been disclosed.

What this says about AI-labelled content in government communications

The fact that a police station itself distributed a manipulated image of an actual operation without identifying it as AI-generated raises questions about transparency in government communications. The arrest was real and the facts are accurate, but the visual representation of it was fictitious. For people who saw the photo, there was no way of knowing that.

In the Netherlands and the European Union, the AI Act imposes obligations regarding the labelling of AI-generated content in certain contexts, particularly when it is intended to realistically depict persons or situations. Thailand falls outside that framework, but the case demonstrates that the need for clear standards governing AI use in government communications is not limited to Europe.

For Dutch and European policymakers and regulators working on the implementation of the AI Act, this incident provides a concrete example of how AI-generated images without labelling can lead to disinformation, even when the underlying facts are accurate. The question of what requirements should be placed on government communications is becoming increasingly practical with cases such as this one.