It’s no secret that artificial intelligence (AI) is slowly creeping into various aspects of our lives. Unfortunately, the same can be said about historical photos that people are starting to fake using this tool. 

While obviously forged images like Napoleon giving a TED Talk or Mata Hari taking a selfie with an iPhone don’t pose any threat and are created for humor and artistic expression, others, which can pass as authentic historical photographs, are crafted to deceive people. 

Just last May, the Facebook account Past Perspectives shared a photo of Henry Ford with his first car in 1896. “Henry Ford sits in his first automobile the Ford Quadricycle. Ford made three Quadricycles in total. Ca. 1896,” the caption writes.

However, the suspiciously clear black-and-white picture doesn’t show Ford or even accurately depict the first car he manufactured. The image also had certain defects, like his right hand seemingly going through the steering wheel, which gave away the image's forgedness.

Even though the image was fake, the page doesn’t provide any information that it’s AI-generated, letting people believe that what they see is an authentic piece of history.

In another instance, the Animals Are Amazing Facebook page shared a photo that looks like it could have been taken in the 19th century, depicting several men standing next to a giant horse. But if one took a closer look, a few things start to raise eyebrows. Like the enormous fingers of a man standing behind the horse or the blurry people’s faces. It was later found out that the image was created as a joke, but at that time it was already shared more than 10,000 times and picked up by other pages as it was real. 

And that’s just the tip of the iceberg. Some people go as far as creating new historical events that never happened and illustrating them with AI-generated images to pass them as reality. Like the “Infamous Blue Plague Incident” of the 1970s in the Soviet Union, the “July 2012 solar superstorm and blackout” in the U.S., or “the 1993 Eruption of Mount Rainier.”

The faking of historical pictures and events was popularized with the creation of a program called Midjourney. It allows anyone to create AI-generated photos from text descriptions. Co-creators Brodbeck and de Barbuat fed content to the machine for several weeks so it could recreate snapshots from the 20th century. The AI-image generator has extremely improved since its first version and now is the most realistic AI model so far.

Experts are highly worried about how this can affect our history and our perception of it. “The potential impact of such posts cannot be overstated. The more these fake images circulate, the harder it becomes to separate fact from fiction. Each new post or share distorts the truth a little bit more, until we're left with a version of the past that bears little resemblance to reality,” says digital colorist and bestselling author Marina Amaral. Not to mention the malicious people who can intentionally manipulate history to fit their own political agenda. 

Even though authorities are trying to solve the issue, the change is happening relatively slowly. In the meantime, what we can do to protect ourselves from historical misinformation is to learn how to identify fake AI-generated images. 

Jo Hedwig Teeuwisse, a historian who specializes in debunking fake historical images, recommends looking at hands. “Most AI programs still have problems with hands, and you often see that people have too many fingers or they just look weird.”

“Then I put them through a reverse image search to try and find when and where they were uploaded first, this often tells us more about the source. I also regularly use AI detection software, it's pretty good although AI creation software is improved all the time so it doesn't always work with newer versions,” Teeuwisse adds.