Artificial intelligence poses risks to election integrity, from fake images to disinformation

From politicians with six fingers to fabricated speeches, artificial intelligence (AI) is increasingly being used in political campaigns, raising concerns about its potential to distort elections, LRT reports.

Lithuanian politicians are turning to AI-generated content to engage voters on social media, but experts and the Central Electoral Commission warn of significant risks. Current regulations lack clear guidelines on AI use in political advertising, a gap that could become more problematic as elections approach.

Linas Petkevičius, a researcher working with AI at Vilnius University, analyzed a campaign by the Democratic Union “For Lithuania” ahead of February 16. The AI-generated images depicted politicians alongside the signatories of Lithuania’s 1918 Independence Act—but with noticeable errors, such as Democrat Lukas Savickas shown with six fingers and Jekaterina Rojaka with four.

“AI algorithms are improving but still struggle to perfectly replicate reality,” Petkevičius said. He warned that AI’s capabilities could be exploited to discredit opponents, such as altering a politician’s speech while preserving their voice, making it nearly impossible to detect manipulation.

Deepfakes and accountability challenges

In Hungary, AI-generated posters ahead of parliamentary elections have already depicted false scenarios, such as Ukrainian President Volodymyr Zelensky and European Commission President Ursula von der Leyen dropping money into a golden toilet.

Lina Petronienė, chair of Lithuania’s Central Electoral Commission, acknowledged that while AI itself is neutral, its misuse poses risks. “AI is just a tool—like a knife, it can be used to spread butter or for harm,” she said. However, she admitted that Lithuania’s legal framework remains unclear on how to handle AI-generated disinformation in campaigns.

Current laws prohibit manipulative election advertising on social media, but enforcement is complex. The commission would rule on whether AI content qualifies as campaign material, while the Radio and Television Commission of Lithuania (LRTK) would assess manipulation—yet both processes are untested in practice.

New labeling rules and enforcement gaps

From August, politicians will be required to label AI-generated content. However, LRTK head Mantas Martišius highlighted another challenge: tracing the origin of harmful deepfakes. “Social media is full of fake accounts, and even if we stop a viral video, identifying the creator is difficult,” he said.

The rise of AI also blurs the line between truth and manipulation. “A real video or recording can be dismissed as fake by a politician claiming it’s AI-altered,” Petkevičius noted, complicating public trust in authentic content.