"Likud" Demands Central Elections Committee Address AI Bias in Elections

The "Likud" party has officially turned to the Central Elections Committee, demanding urgent clarifications regarding measures to address political bias in artificial intelligence models ahead of the upcoming 26th Knesset elections.

In a formal letter sent by attorney Ilan Bombach to committee chairman Judge Noam Solberg, the party requested comprehensive information concerning steps being taken to counter research findings that reveal political bias and data poisoning in AI language models.

Research Findings on AI Bias

The inquiry follows extensive reporting detailing how conversational AI models influence Israeli voters. The coverage highlighted recent studies from the Institute for National Security Studies (INSS), indicating that popular language models are vulnerable to political leaning and external manipulation.

Research conducted by Dr. Ofer Friedman, examining platforms such as ChatGPT and Gemini, revealed a center-left inclination when asked to evaluate Israeli parties based on national security, economy, and social cohesion. Both systems ranked Gadi Eisenkot's party at the top, while placing "Likud" in sixth place, and positioning right-wing factions at the bottom.

"This is a particularly severe bias given the expected scale of AI tool utilization during the election campaign," attorney Bombach stated in his letter.

Calls for Committee Action

Experts emphasize that the observed bias is an inherent technological phenomenon rather than a deliberate conspiracy to influence elections. However, findings from another study by Dr. Shay Hershkovitz demonstrated how easily manipulated data can be planted online to influence AI outputs.

"The upcoming elections are already compromised from this perspective given the tight timelines and entirely new paradigms," noted David Siman-Tov, a senior INSS researcher.