Forum for Democracy Party Leader Thierry Baudet retweeted Twitter accounts with fake profile photos several times in the past month.
How exactly can you spot those fake profiles?
Baudet retweeted a message from the account @MeneerPanini with the name Eduard van Vlugtenstahl on April 15.
"Had an annoying post today," the user wrote.
"My ex-father-in-law died suddenly after a second shot."
A second retweet followed on May 5, this time a message from @ PietVeenstra4.
"When I look at all political parties, you are the only ones who stand up for us", Veenstra Baudets colleague Pepijn van Houwelingen assured.
Both Twitter accounts seem harmless at first glance.
They are constantly posting messages and have a realistic looking Twitter biography.
Van Vlugtenstahl even links to the Twitter account of his cousin @gerrisma, who also confirms on his profile that he is family.
The reality is different, observed assistant professor and fact-checker Peter Burger.
According to him, these are profiles with computer-generated faces of people who do not really exist.
Baudet retweets another radical right troll account with a computer-generated face.
Avatar Author JPeterBurger Moment of places 5:41 - 6 May 2021
Software can generate random faces
Special computer software is able to generate random faces.
Artificial intelligence is used to look at a huge amount of photos of real people and then automatically take a new photo based on the acquired knowledge.
Chip manufacturer NVIDIA has already shown how realistic that can look with its software StyleGAN, which can be downloaded for free to take a photo yourself.
Since then, the software has been widely misused for profile pictures on fake accounts.
Online trolls generate dozens of such images to fill multiple accounts at the same time and thus pretend to be a larger group.
This makes it appear that many people have a specific opinion, when in reality it is spread by a small group.
This is how you recognize these fake photos
On the Sensity website, you can scan photos to see if they were generated by a computer.
This site was founded by Giorgio Patrini, who previously conducted research at the University of Amsterdam into machine learning, the technique used by this software.
He also researched deep-fake porn videos with a previous company.
During a scan by the site you will see how likely the image is to be manufactured and which software is likely to have been used.
You can scan five images a day, after that you have to pay.
The photos from Baudets retweets are both fake, according to the site.
At @ PietVeenstra4 there is a 71.8 percent chance that his photo was taken with the computer program PGGAN.
@MeneerPannini was definitely shot with StyleGAN2 for 99.8 percent, just like the photo of the said cousin.
According to Burger, you can often recognize synthetic photos by the position of the eyes, which are always in the same place of the image.
This is not conclusive evidence, but it is indicative: it is a way to spot a possible fake profile, after which you can analyze the image to be sure.
@mterburg @JPeterBurger What you often see with generated photos is that the eyes are (almost) always in the same place.
Avatar Author Kootstra Moment of places 06: 10 - 6 May 2021
Account owners are unknown
It is not clear who is behind the Twitter accounts in question.
Other tweets from the profiles are full of fake news and untruths about, for example, the corona vaccine, which according to the Twitterers would be deadly and dangerous.
Fake accounts are nothing new on Twitter, by the way: trolls tried to disrupt the political process in this way as early as the American elections in 2016.
At the time, this usually happened from Russia.
Twitter has since been trying to track down and suspend fake accounts.
In 2018, the number of users on the platform therefore decreased by nine million.