Israeli Research Warns Nation Unprepared for AI-Driven Election Manipulation

A comprehensive study by the Knesset Research and Information Center, authored by Roy Goldschmidt and Dr. Yael Levi Ariel, highlights the complex challenges Israel faces ahead of upcoming elections due to advances in artificial intelligence (AI). The report reveals that AI has transformed political influence from an expensive, large-scale operation into an accessible, automated, and low-cost process. Previously requiring massive budgets and organized troll farms, now open language models can generate thousands of fake Hebrew-speaking accounts that convincingly debate on social media and spread personalized fake news.

The evolution from simple spam bots to sophisticated chatbots based on large language models from OpenAI, Google, and X enables short, persuasive conversations that can shift voters’ political views more effectively than traditional TV ads. The threat extends beyond deepfake videos and audio to data poisoning, where manipulative falsehoods are fed into AI models, which then replicate these lies as objective facts. While Israel relies on a 1959 propaganda law to combat these threats, Western countries have adopted more advanced regulations. The European Union, for example, enforces strict transparency and labeling requirements under the AI Act and Digital Services Act, leading companies like Meta and Google to halt political ads entirely in the EU.

In contrast, the United States struggles with federal legislation due to First Amendment protections, leaving states such as New York, California, Minnesota, and Maryland to enact laws banning political deepfakes near elections or requiring disclosure. China employs strict centralized regulation to prevent unauthorized online manipulation, though elections are less relevant there due to single-party rule.

Recent global incidents demonstrate the real and immediate danger of AI misuse in politics, including voice cloning of the U.S. president to suppress voter turnout, deepfake videos of candidates in Ireland and Germany announcing withdrawals, and election result annulments in Romania amid suspected Kremlin-backed interference. This environment fosters public distrust, enabling politicians caught in scandals to dismiss authentic recordings as AI fabrications, a phenomenon termed the "liar’s dividend."

In Israel, responsibility for countering malicious influence is fragmented among the Central Elections Committee, Shin Bet, police, the Cyber Department of the Attorney General’s Office, and the National Cyber Directorate. A new law effective as of Sunday mandates clear labeling of any election propaganda created or significantly edited using AI, covering visual and audio content. However, with foreign and domestic actors wielding advanced digital brainwashing tools, Israeli democracy must do far more than rely on algorithms to safeguard election integrity.