Are these photos of Trump being arrested?

We came across the following post on Twitter, which had 7.2 million views at the time of writing:

The tweet, posted on 21 March 2023, the same day that Trump himself predicted he would be arrested, seems to show the former President being apprehended by police officers  and placed under arrest.

Aside from the heavily stylized nature of the photos, the images seem realistic and very plausible, given the timing of the posts and the previous prediction of arrest by Trump. When we conducted a reverse image search on the pictures however, we found that the pictures had been created by Eliot Higgins, founder and creative director of investigative collective Bellingcat and captioned: “making pictures of Trump getting arrested while waiting for Trump’s arrest”.

In the Twitter thread posted by Higgins, he explained that he used the AI engine Midjourney, now in its 5thiteration to generate the realistic images. Midjourney, an American artificial intelligence programme, generates images from natural language commands or prompts. According to Higgins, he used the prompt “Donald Trump falling over while getting arrested. Fibonacci Spiral. News footage” in order to generate the images seen in the original tweet.

The claim that these photos show former President Donald Trump being arrested by law enforcement in Washington DC is therefore false. The images were generated by an artificial intelligence text-to-image generator called Midjourney.

These AI-generated photorealistic images along with discourse surrounding how large language models might be misused for disinformation purposes point to the serious threat that AI could pose to the information environment. With the ability to generate highly convincing text and images, these models could be used to create false stories that are difficult to distinguish from real news.

 It is thus important to consider the risks and challenges associated with these models. By taking a responsible and cautious approach to their development and use, it is possible to mitigate the risk of AI-generated disinformation, while being able to benefit from their potential in shaping journalism and improving fact-checking capabilities.

Leave a Reply