As AI Blurs Fact and Fiction, Israel’s Election Committee Faces New Tests

Jerusalem, 21 October, 2026 (TPS-IL) -- Israel’s Central Elections Committee is being thrust into an unlikely role ahead of the next election: refereeing disputes over artificial intelligence, political propaganda and fabricated content that can make events that never happened look real.

Ahead of the Oct. 27 elections, the committee has introduced labeling requirements for AI-generated election propaganda, while its chairman, Supreme Court Justice Noam Sohlberg, has ordered the removal of videos deemed misleading. But experts told The Press Service of Israel that enforcement has limits, and other democracies face similar challenges as fabricated content and coordinated online campaigns complicate efforts to distinguish genuine political information from manipulation.

Central Election Committee chairman Justice Noam Solberg inspects ballot slips at a printing plant in Kiryat Gat on Oct. 5, 2026. Israel’s elections are scheduled for Oct. 27. Photos by Yossi Zeliger/TPS-IL

“The problem lies not in the use of the technology itself, but in the attempt to conceal it,” Aviad Nimekovski, deputy general counsel of the Central Elections Committee, told TPS-IL. He described political persuasion as fundamental to democracy but said problems arise when campaigns attempt to mislead audiences.

Israel introduced a temporary legislative amendment requiring election propaganda created using AI to carry an explicit label identifying it as artificially generated. The committee has also established a dedicated team to study AI and other forms of potentially illicit political influence, while maintaining contact with social media companies including Facebook, Microsoft and Google.

“We are among the first in the world to implement such a measure,” Nimekovski said, adding that its effectiveness would be evaluated over time.

When Anyone Can Make a Fake

The committee can issue injunctions within minutes or hours, but removing a video does not prevent it from being reposted elsewhere. Nimekovski said the committee receives around 250 petitions, most concerning election propaganda, from across the political spectrum.

The ease of producing such material adds to the challenge. Jonathan Hasidim, CEO and co-founder of Scooper, a Bnei Brak-based social media monitoring and public opinion analysis company, said AI has sharply reduced the resources needed to create political videos.

“Nowadays, you no longer need a film crew, actors, specific locations, or large production budgets to create a convincing political video,” Hasidim told TPS-IL. “All that is required is an idea, a computer, and a few minutes of work.”

Campaigns can depict opponents meeting unpopular figures or forming damaging alliances, even when such events never occurred. Some videos are obviously fictional; others closely resemble authentic footage.

Aya Yadlin, a senior lecturer at Bar-Ilan University’s School of Communication, said even openly fictional videos can influence perceptions by portraying political opponents in degrading or threatening scenarios.

“At the openly fictional end, parties have released GTA-style parody videos casting rivals as video-game characters,” Yadlin told TPS-IL. “The main risk with this kind of content is not deception but rather that rivals are dehumanized.”

More realistic fabrications pose a different problem: viewers may mistake invented events or relationships for fact. In July, Sohlberg ordered Finance Minister Bezalel Smotrich, leader of the Religious Zionism party, to remove a video depicting Yashar! party leader Gadi Eisenkot shaking hands with Palestinian leader Mahmoud Abbas.

“The segment in the video presents an event that did not occur but appears realistic. It shows a political connection between Eisenkot and Abbas, a connection that does not exist, and therefore may mislead on this matter,” Sohlberg ruled.

The Limits of AI Labels

The problem extends beyond AI-generated content, Yadlin said. She pointed to “fauxcasts,” campaign videos staged to resemble journalistic podcasts, complete with studios and microphones but without an actual interview.

“The fauxcast does not lie in its words. It lies in its set, and it exploits the trust audiences have built in the podcast format,” she said.

Israel’s Elections (Propaganda Methods) Law dates to 1959, long before social media and generative AI. While AI-generated campaign content must now be labeled, Yadlin noted that similar requirements do not cover every misleading format that imitates journalism.

Labels themselves are an imperfect safeguard. Videos can be cropped or reposted without their original disclaimers.

“Labels help, but less than people expect,” Yadlin told TPS-IL. “In US-based experiments, AI labels tested reduced belief in the claims made in the fake content, but labels simply stating content was AI-generated had little effect on people’s intention to engage with the post. We must understand that simply labeling something as ‘AI’ is not similar to mandating a ‘false’ label.”

Hasidim called for a broader response combining legislation, monitoring, enforcement and public education. He also warned that no detection system can reliably identify every AI-generated image, video or voice as the technology improves.

“The requirement to label content as ‘AI-generated’ is akin to trying to cure a terminally ill patient with a simple painkiller,” he said. “In an era where anyone can easily create AI-based content, regulation that applies only to politicians is far from sufficient.”

Authorities should therefore examine not only whether content appears artificial but also who distributes it, how quickly it spreads and whether accounts are coordinating their activity, Hasidim stressed.

From Deepfakes to Coordinated Campaigns

Israel’s experience reflects a wider challenge for democracies, although the methods vary. In the United States, an AI-generated imitation of President Joe Biden’s voice was used in a 2024 robocall that urged New Hampshire voters not to participate in the Democratic presidential primary. In Slovakia, a fabricated recording circulated just two days before the 2023 parliamentary election appeared to show a candidate discussing election rigging with a journalist.

Romania’s 2024 presidential election highlighted a related problem: coordinated online activity can amplify political messaging even when the content itself is not fabricated. Yadlin cited a network of roughly 25,000 synchronized accounts that helped bring a little-known candidate to prominence on TikTok, illustrating how apparently independent online activity can be orchestrated at scale.

The common thread is not simply the production of fake videos. Cheap generative AI can supply convincing content, while networks of accounts and influencers amplify it, making coordinated propaganda appear to be spontaneous public opinion.

Yadlin said authorities should focus on identifiable actors, requiring official campaigns to disclose and take responsibility for their material at the source. Social media companies could introduce additional checks on sharing and dedicate specialized teams to election-related content.

“The structural problem is timing. Rulings arrive after the views have already accumulated,” she said of the Central Elections Committee’s role.

The risks extend beyond fabricated material. Hasidim warned that politicians could invoke AI as an excuse to dismiss authentic evidence as manipulated, a phenomenon known as the “liar’s dividend.”

Yadlin said the resulting erosion of trust could be more damaging than individual falsehoods. “Democracy depends on a shared factual baseline,” she said. “Voters can only hold leaders accountable if they can agree on what actually happened.”

Israel’s efforts to regulate AI-generated election propaganda may offer lessons for other countries, Nimekovski said. The committee has studied international cases, run public awareness campaigns and consulted technology platforms on enforcement.

But the challenge, the experts said, is to address deception without treating AI itself as illegitimate or assuming that suspicious-looking material is necessarily false.

“The result is a public that believes nothing rather than one that believes lies,” Yadlin said. “A public like that is easier to demobilize and harder to inform.”