How AI Is Reshaping Election Campaigns in Israel

TEL AVIV—The Israeli prime minister appears asleep in bed, while warnings of a Hamas attack flash across the screen. Senior judges are shown being dragged through the streets after protesting the government. The mayor of New York holds a party to celebrate the downfall of longtime Israeli leader Benjamin Netanyahu.

In the corner of each piece of footage a small message carries the label “AI” and in Hebrew reads: “The content was produced by digital means and wasn’t actually documented.” Such footage has been posted online by the country’s most senior lawmakers, as well as the prime minister.

This is the era of AI-generated campaign videos, a phenomenon evolving with such ferocity that it makes traditional election attack ads look archaic—and tame.

Israeli politicians are churning out and promoting AI content to discredit their rivals ahead of a crucial fall vote being cast as the first general election of the AI age. Regulators can’t keep up.

The poll is the first election in Israel since the Hamas-led attacks of Oct. 7, 2023, and the subsequent war in Gaza, and is widely seen as a referendum on Netanyahu’s handling of one of the most tumultuous periods in the nation’s modern history.

Even as warnings proliferate about the existential threat posed by the technology, AI content is being produced faster and more cheaply than ever, making the Israeli vote, set for Oct. 27, a laboratory for its impact on the democratic process, say AI researchers and developers.

Israel’s election regulations, which date from the 1950s, address radio and print newspapers, but never imagined the speed misinformation could spread when propelled by AI. While regulations have been temporarily updated in each election, researchers say they aren’t robust enough to match the lightning pace of the latest technological advancements.

To prepare for potential threats, Israel’s election committee has conducted dozens of wargames, including with the police and the Shin Bet internal security service. One scenario imagined that AI content was published on the day of the election claiming that voting stations have closed in one city due to a security crisis. The committee declined to give details on the simulations or their outcome.

Israel has one of the world’s fastest rates of AI adoption among its citizens, according to international surveys and Israel’s internet association.

“The pace of progress of AI makes it actually the first real AI elections in the world,” said Tehilla Shwartz Altshuler, a senior fellow at the Israel Democracy Institute who specializes in technology law and policy.

The elections are also seen as the most important in the nation’s history by many in Israel, where opinions are divided over Netanyahu, who has opposed setting up an independent state commission to investigate whether political failures contributed to the Hamas-led attacks. Israel is also facing growing international isolation over the war in Gaza, worrying many Israelis.

Roughly one in eight Israeli election-related social-media posts published in recent months were created or altered with AI, according to Ze Be AI, a research initiative by the Open University of Israel.

Netanyahu’s hard-right Likud party is behind 46% of all the AI posts, according to the initiative.

“This is one of the most challenging elections campaigns we have ever faced,” said Dean Livne, acting director general of Israel’s Central Elections Committee, who has overseen nine elections on the committee. “You suddenly have a new technology that is amazing, which allows you to analyze information quickly,” he said. “But also to deceive, create images which don’t exist in reality with little effort.”

With AI reducing the cost of producing expensive election videos, politicians are motivated to produce AI-generated content even more, said Yonatan Zinder, chief executive of Blueye AI, an Israeli startup.

“It’s not a game of quality but quantity,” he said.

In addition to challenges posed by AI content, law enforcement agencies fear domestic and external influence campaigns turbocharged by AI.

The technology has made it easier to create more bots and fake users and more high-quality content compared with just a year ago, increasing the scale and reach of influence campaigns, said Itai Yonat, founder and CEO of Intercept 9500, an Israeli company that focuses on defenses against such campaigns.

A report by Facebook parent Meta Platforms last month found that an AI network operating out of Israel used AI agents to generate and disseminate, at scale, comments that were tailored to their context to make them appear more genuine. This type of operation suggests AI can now be used to “produce, adapt and distribute content with minimal human interaction,” the report said.

The election committee has established a new team focusing on influence campaigns in coordination with the country’s security services, amid concerns about attempts to meddle in the heated elections by foreign countries such as Iran and Russia.

The committee itself has been a target of a domestic attack campaign that began in March, which called into question its legitimacy, according to Intercept 9500. Genuine pro-Netanyahu media personalities took part in the campaign, but so did bots and bot-like accounts, which represented 17% of the posts, the company said, fueling concerns about potential efforts to call the results of the election into question.

During a previous election round in 2015, Netanyahu falsely claimed that “Arab voters are coming out in droves to the polls. Left-wing organizations are busing them out,” to boost turnout for his right-wing party and was criticized for delegitimizing the votes of Arab citizens of Israel. Netanyahu later apologized for the statement.

Researchers worry that AI usage in the polls could reach a crescendo on election day itself, for example through increased AI-powered robocalls made to sound like real politicians repeating misinformation.

Political parties and candidates are also working to manipulate the results provided by chatbots to produce answers in their favor. At least one politician from the Likud party is currently hiring a consultant for generative engine optimization, to improve the results given by chatbots to questions about him, according to messages viewed by The Wall Street Journal.

A new law announced by the election committee in late July requires all election propaganda created with AI, including photos, videos and audio, to be clearly labeled as such. Potential violations of the law can be flagged to the committee via a petition for a ruling. The petition process is time-consuming, and researchers say that by the time a video is taken down, the damage is already done.

A live dashboard developed by the Brandeis Institute, an Israeli think tank, monitors political accounts on Israeli social media, and analyzes tens of thousands of posts for potential election law violations. The dashboard identified 166 undisclosed AI-generated posts last month. The election committee has so far ordered the removal of only a handful of AI videos that violate its rules.

An estimated 98% of content flagged by the dashboard as violating AI content rules doesn’t even reach the election committee, according to the institute.

The election committee says it doesn’t actively monitor social media for potential violations and that it considers violations when they are brought to the committee’s attention as part of a petition.

A video posted by National Security Minister Itamar Ben-Gvir showed AI-generated overweight Palestinian detainees on a conveyor belt to an Israeli prison, from where they emerged thin and distraught. Ben-Gvir deleted the post following criticism, but not before it amassed millions of views online.

The central election committee will have to adapt to working in terms of hours, if not minutes, said Shwartz Altshuler. “Because on election day, when we will see AI-generated fake voice messages, or when we will talk about a campaign to discredit the outcome of the election, they won’t be able to wait two weeks until they have a ruling.”