AI Search
It’s a stark contrast to the approach taken in Washington, where U.S. President Donald Trump has frequently used AI-generated content to get his point across.
AI generated Text-to-speech
BRUSSELS — The European Union’s main institutions have banned staff from using artificially generated videos and images in official communications.
Amid growing scrutiny of the use of artificial intelligence and deepfakes online, the European Commission, the European Parliament and the Council of the EU all have policies in place that bar their press teams from using fully AI-generated visuals, they told POLITICO.
It’s a marked contrast to the approach taken in Washington, where U.S. President Donald Trump has frequently used AI-generated content to get his point across, including attention-grabbing posts such as a madcap AI-generated video on his ambitions for the Gaza Strip.
Several EU governments are also deploying the technology in their communications, ranging from posts by the German chancellor to teach people about artificial intelligence to deepfake videos from the Hungarian prime minister attacking Brussels.
While Brussels’ stance is designed to protect the credibility of messaging, it is also raising questions about the EU’s ability to remain relevant in an era of political communications in which the creative use of AI is on the rise.
The EU is “rightfully thinking first and foremost about risks: whether their content may be perceived as deceptive or harmful, whether it impersonates reality, and whether they can offer clear accountability and disclosure,” said Alexandru Voica, head of policy at the U.K.-based video generator Synthesia.
But several questioned whether an outright ban makes the most sense as fast-moving geopolitical crises raise the stakes for institutions to have the strongest possible presence online.
The risk that deepfakes could erode trust “should not paralyze you into doing nothing,” said Renout Van Zandycke, who runs the Belgian communications agency Exposure, which has advised several political parties. “Doing nothing is equally as bad.”
“Responsible use beats abstinence,” said Walter Pasquarelli, an adviser to the OECD who also researches AI-generated content at the University of Cambridge.
Some estimates suggest that AI-generated content has now overtaken the amount of human-generated content online, with 8 million deepfakes shared online last year.
Political communication is not immune. As hostile deepfakes have rocked election campaigns from the Netherlands to Ireland, those holding office have also picked up on the practice.
Donald Trump has used AI in 36 posts on his Truth Social account since his inauguration, the Poynter Institute, a global fact-checking non-profit, said in October. Among them were an image of him as the pope and a video of him dumping feces on protestors.
The EU is intentionally taking a different approach.
“The footage and photos that we are using and making available for journalists’ use or for official information purposes do not include AI-generated content,” European Commission spokesperson Thomas Regnier told POLITICO.
He said the Commission aims to “foster citizens’ trust” and, for that reason, “authenticity” is a priority in its communications. The Commission does allow the use of AI to optimize footage, such as to enhance image quality, he added.
The European Parliament has guidelines for its staff on the use of generative AI tools “emphasizing vigilance regarding inherent risks,” a spokesperson said.
The EU’s stance reflects a genuine fear that the use of AI images and videos is causing confusion online, as the line between real and fake gets increasingly blurred.
“In an era of eroding trust, there’s a genuine argument that the most powerful thing an institution can do is to refuse to play the game,” Pasquarelli said.
Yet by snubbing AI-generated content, the Commission also risks falling behind in an era where diplomacy is increasingly conducted online.
“How quickly and effectively you respond is now becoming more important than ever,” said Voica from AI company Synthesia.
A much-cited example is the French Response account on X, through which the French Ministry of Foreign Affairs frequently claps back against false takes or accusations from the accounts of foreign adversaries.
Brussels’ choice to sidestep deepfakes is also seen as a potential missed opportunity to educate people about AI-generated content, where the EU looks to be a front-runner. Under the bloc’s AI law, AI-generated content must be watermarked and labeled to make it recognizable.
The Commission should make it clear that synthetic content in itself is not a problem, said Pasquarelli — whereas undisclosed synthetic content is.
“By refusing to engage with it altogether, the European Commission is missing a leadership opportunity to demonstrate what responsible, transparent use of AI in political communication actually looks like,” he said.
“If public institutions never use these tools in visible, responsible, well-labeled ways, they miss an opportunity to help people understand what AI looks like in practice,” Voica said.
German Chancellor Friedrich Merz recently did just that — posting a fake dancing version of himself in an Instagram clip to warn about AI risks and opportunities.
Regulators complain of surging workload and bottlenecks as citizens turn to AI to interact with officials.
The tech company made the announcement after finding that its new Claude Mythos model outperformed most humans at “finding and exploiting software vulnerabilities.”
‘If you don’t have artificial intelligence in your systems, you actually don’t have an army,’ says Mistral CEO Arthur Mensch.
Karin Keller-Sutter sued an “unknown” perpetrator for prompting the chatbot to come up with offensive comments.
