Analysis: Israel's Election Faces a Bigger AI Threat Than Deepfakes, Researchers Warn | Israel Politics Post
Analysis: Israel's Election Faces a Bigger AI Threat Than Deepfakes, Researchers Warn
While regulators focus on labeling AI-generated campaign videos, experts say the real danger is chatbots quietly feeding voters manipulated answers about candidates and parties, a risk Israel's rules barely touch.
3-MIN READ
By Israel Politics Post Staff·Published
A Likud campaign stop in the southern Israeli city of Ashdod · Source: Liron Moldovan/Flash90, via The Times of Israel
Israel's October 27 election could become one of the world's first major votes genuinely shaped by artificial intelligence, and the biggest risk may not be the one regulators are focused on, The Times of Israel reported, citing researchers who study the technology's political impact.
Much of the public conversation has centered on deepfakes. Opposition leader Yair Lapid has warned that voters could be shown convincing AI-generated video of him falsely claiming he would form a government with Arab parties. Israel's Central Elections Committee responded in July by introducing labeling requirements for AI-generated campaign material. But researchers argue that threat, however real, is also the easiest for voters to eventually question once they know to look for it.
The harder problem, according to researcher Shay Hershkovitz, is what happens inside the chatbots nearly 90 percent of Israelis already use, per a May survey from the Israel Internet Association. Techniques that seed information online specifically to shape how AI systems answer political questions can quietly bias a chatbot's response on a candidate or party, with no visible manipulation for a user to catch. "The poisoned retrieval layer is a problem you cannot" easily detect, Hershkovitz said, contrasting it with a deepfake video a viewer can at least scrutinize.
Romania's annulled 2024 election is the cautionary example researchers keep returning to, where a previously obscure far-right candidate surged through coordinated AI-generated content and bot activity before the results were thrown out entirely. Israel's current rules do not meaningfully address that kind of chatbot-level manipulation, and the country's comparatively light regulation of tech platforms, with companies facing what the Internet Association describes as "no accountability whatsoever" compared to European rules, leaves voters more exposed than in many other democracies heading into a close and consequential vote.
