One analysis says it is difficult to assess how artificial intelligence (AI)-generated imagery used by far-right groups influenced France’s first parliamentary elections and the European Parliament elections.
advertisementFar-right parties in Europe are using artificial intelligence (AI)-generated images to spread their political messages, two new reports say, but it is difficult to assess what impact these images may have had on recent French and European elections.
The nonprofit organization AI Forensics examined the social media pages of 31 French political parties between May 1 and June 28, during the EU elections and the first round of the French legislative campaign, to see if they could uncover generative AI (genAI) content.
They then ran the images through AI-related search engines to confirm they had been created by AI tools.
In total, the team found 51 examples of generated AI imagery used by French political parties, with the far-right parties Rally National (RN), Reconquiete and Le Patriot being most frequently involved, though the researchers said this was a conservative estimate.
“This is the first election cycle where we see (generative) AI, but I’m convinced that we will see it being used more and more in the coming years,” Salvatore Romano, a researcher in AI forensics, told Euronews Next.
“If we can’t manage this now at scale, it will be even harder to manage when the technology becomes indistinguishable from the actual content,” he added.
The far-right RN party won the first round of the French elections last Sunday with 33% of the vote, and also won in June’s European Parliament elections with 31.5%.
Where AI content comes in
According to the AI Forensics report, most of the AI-generated election campaign images used by French political parties “convey anti-immigration, unsafe France, lack of freedom of expression, flaws in the judicial system and insufficient agricultural protection messages.”
“[The images] “It’s being used as a scare tactic to make people feel that France is not safe within the EU, and it’s very targeted at vulnerable members of society,” added Miasia Schuler, an AI forensics researcher.
Schuler said the images often look “hyper-realistic,” and it’s clear to the average user that they’ve been rendered by AI.
Others were more subtle, such as imitating political cartoons that appeared in newspapers.
Some parties, such as the far-right Reconquisté party, have also used generative AI to create images of French President Emmanuel Macron and European Commission President Ursula von der Leyen.
According to an investigation by the Atlantic Council’s Digital Forensic Labs (DFRLabs), the RN launched a website ahead of the EU election campaign called “A Europe Without Them”, which similarly challenged the leadership positions of Macron and von der Leyen.
The website, which is still up and running, contains at least three AI-generated images displaying petitions against “EU disrespect for people, mass immigration and the promotion of the hijab,” the report found.
“This is just the tip of the iceberg.”
Valentin Châtelet, a researcher at the Atlantic Council, said it was “difficult” to know how much impact this kind of AI-generated content had on the RN’s success in the EU and French elections.
“I think this has had a positive impact on the materials the party is using to communicate certain ideas, such as Mr. Bardella’s particular position on certain issues,” Châtelet said.
“[AI-generated]images will also be used to shock people and eventually become mainstream,” he added.
Euronews Next contacted the National Union (RN), Identity Democracy Group, Patriots and Reconqué parties but did not receive a response.
advertisementAccording to the French broadcaster, RN candidates will receive AI training starting in January 2024, including how to use ChatGPT. movieThe broadcaster said the training booklet provided examples of how AI could help candidates, such as creating election posters or finding material for debates.
Neither the AI Forensics investigation nor the Atlantic Council’s DFRLabs looked at candidates’ personal posts or content created by fans of political parties.
Romano believes the research showing how much unlabeled AI-generated election content is circulating on social media platforms is “just the tip of the iceberg.”
Tech companies’ “gross negligence” in using AI
The AI forensics report found that none of the AI-generated images were labeled as such.
“Our findings reveal a serious failure by both political parties and party platforms to uphold their commitments to label AI-generated imagery in political campaigns, highlighting the urgent need for transparency,” the report said.
advertisementIn February, major tech companies including Google, OpenAI, Meta and TikTok signed an open technology agreement “setting out expectations for how signatories will moderate deceptive AI-generated election content” on their platforms ahead of the European elections.
The first promise of the agreement is that platforms will be able to “identify (realistic AI-generated images) and verify the authenticity and origin of the content.” Accord Read.
The company also promised to review its models to understand the risks of deceptive AI content.
Several political parties signed off on their own versions of this. agreement The European Commission is “committed to combating deliberate misleading of citizens, including through the use of artificial intelligence,” the statement said.
The European Parliament’s Identity and Democracy group (ID), of which the RN is a member, signed the code of conduct. No other national French parties did.
advertisementRomano said AI Forensics plans to submit a report to the European Commission in the hopes that some retaliatory action will be taken against parties that violated the code of conduct.
