Israel’s AI Election Labels Arrive Too Late for Shadow Campaigns

An AI-generated Likud campaign clip showed Gadi Eisenkot veering from a young man in an “Israeli unity” shirt to embrace Ra’am leader Mansour Abbas. Critics said the youth resembled Eisenkot’s son Gal, killed in Gaza in December 2023. Likud denied intent and swapped the figure for a woman. Days later the Central Elections Committee ordered Bezalel Smotrich to delete another AI video that invented political meetings. Temporary labeling rules took effect July 26. They address official ads. They do not close the larger gaps.

Israel heads into its first Knesset election since October 7 with AI tools cheap enough for any activist or foreign desk to fabricate convincing video, audio and images. The election is set for October 27, 2026. Labels on party material are a start. The second-order effects sit elsewhere: foreign networks that never label, surrogate accounts that parties can disown, and platforms that now write their own election codes faster than the state.

Two Clips That Made the Rules Real

The Likud video landed on Netanyahu’s X account in mid-July. Eisenkot, former IDF chief and now a leading opposition figure, runs through a sunlit field toward a young Jewish man before turning at the last second to hug Abbas. Outrage focused on the resemblance claim and the cruelty of using a fallen soldier’s image, even by implication. Likud revised it. The damage was already done.

Separately, Smotrich’s Religious Zionism party posted an AI clip that placed Eisenkot shaking hands with Mahmoud Abbas and Mansour Abbas, implying support for a Palestinian state he has not endorsed. Eisenkot petitioned. On July 26 Justice Noam Sohlberg, Central Elections Committee chairman, ordered the video removed from all party platforms. He ruled it depicted events and relationships that never occurred and could mislead voters. A simple disclaimer was not enough; Facebook users often miss labels in the feed layout.

  • Likud clip: unity figure swapped after backlash over Gal Eisenkot resemblance claim
  • Smotrich clip: fabricated meetings with PA and Ra’am leaders, ordered deleted
  • Both used realistic AI that could pass as authentic footage at a glance
  • Sohlberg stressed deception risk over artistic intent

These were not abstract warnings. They were the first major AI tests of the new campaign season, and they arrived before the labels were even mandatory.

What the Disclosure Rules Demand

The temporary regulations rest on an amendment to the Elections (Propaganda Methods) Law. They cover material first published on or after July 26 that is digitally created or substantially altered in ways that could be mistaken for authentic content of a person, place, event, document or object.

A disclosure notice can be text stating the content was created or materially edited by digital means, or one of two standardized symbols in the rules appendix: one for wholly or predominantly synthetic material, one for material alteration. detailed formatting rules for AI disclosure notices require high-contrast plain backgrounds, prominent placement, legible non-condensed fonts, and the same language as the content.

Medium Placement and Size Key Extra Rule
Print / still image At least 5% of height or 10-point font, whichever larger Prominent position, high contrast
Video Visible for entire duration of altered content; font ≥5% of screen height or width Must stay on screen throughout
Audio only Clear reading at beginning and end Volume and speed match the rest of the piece

Compliance with the notice creates a presumption that the publisher met the statutory disclosure duty. The rules do not replace other accuracy or labeling laws. Not every AI use triggers a label; only content that could fool a reasonable viewer into thinking it is real documentation.

Failure to disclose can be treated as a violation. That is the enforcement hook the editorial wants used quickly and with real teeth.

Foreign Networks Sit Outside the Label Box

President Isaac Herzog has warned that hostile actors inside and outside Israel aim to deepen internal divisions during the vote. Senior security officials told The Jerusalem Post that Turkey is running social-media activity meant to shape opinion and widen rifts, placing it alongside Iran. They did not name accounts or confirm direct government direction.

The State Comptroller special report on foreign digital influence, published in July 2026 and reflecting the situation as of March, found no comprehensive national policy and no single authority designated to lead the inter-agency response. Preparations routinely begin only shortly before elections. Monitoring gaps, low public awareness, and siloed agencies leave the country exposed. Artificial intelligence multiplies the problem by letting adversaries generate volume and realism at low cost.

An INSS analysis of the conceptual failure argues the deeper issue is the lack of an agreed national concept of foreign information manipulation and interference. Intelligence, strategic communications, justice, cyber and education each own a slice. None owns the whole. Shin Bet treated the threat more seriously only in January. The National Security Council had largely stepped back. A new coordinating body under the Prime Minister’s Office is recommended, but structure without concept will recreate the same gaps.

Labels on Israeli party channels do not bind Iranian, Turkish or anonymous networks. They do not stop coordinated inauthentic behavior that amplifies real domestic grievances. That is the second-order exposure the rules leave open.

Platforms Write Their Own Playbooks

While the state debates coordination, TikTok moved first. The company created a dedicated Israeli election task force of cybersecurity, deceptive-behavior and integrity specialists. It draws on experience from more than 250 elections worldwide. Tools include automated detection, human review, partnerships with the Central Elections Committee and Reuters as fact-checker, invisible watermarks on its own AI tools, and provenance standards that can auto-label content created elsewhere.

Key TikTok measures for the Israeli race:

  • Realistic AI content must be labeled; impersonation of candidates, parties, the CEC or media is banned and removable even if labeled
  • Paid political ads and creator payments for promotion are prohibited; government and party accounts cannot monetize or boost
  • False procedural information about voting dates, locations or eligibility is barred regardless of intent, including jokes
  • Covert influence ops (bots, fake engagement, coordinated amplification) are targeted; the company reported removing more than 40 such ops globally in the first half of the year
  • In-app Election Center will carry CEC-sourced voting info and disinformation tips; unverified posts can be down-ranked or labeled

TikTok removed more than 551,000 videos in Israel in the first quarter of 2026 alone. Other platforms face the same pressure. The editorial is right that they should act fast on clear fabrications of candidates. Yet the result is private companies setting the practical boundaries of Israeli political speech at digital speed, while the state still lacks a permanent foreign-influence lead.

Crowd discussion on X noted the task-force announcements as platforms becoming de-facto election police. That shift is already visible. It fills a vacuum the Comptroller documented. It also means enforcement standards can differ by company and can change without Knesset debate.

Who Owns the Surrogate Feed

The Jerusalem Post editorial draws a clean line: parties must answer for official channels and for coordinated content distributed on their behalf. A campaign should not harvest the benefit of a fabricated video then blame “supporters” or unaffiliated pages. That principle is easy to state and hard to police.

Anonymous Telegram channels, WhatsApp forwards, TikTok clones and X accounts can circulate the same deepfake within minutes. By the time a complaint reaches the committee and a ruling issues, the clip has already shaped the next day’s conversation. Sohlberg’s Smotrich order showed he will act when the material is clearly on a party platform and clearly misleading. The harder cases are the ones designed to stay one hop away from official responsibility.

Every camp faces the same temptation. AI lowers the cost of volume and emotional precision. Coalition and opposition, Jewish and Arab lists, will all be tested. Equal application of the rules is the only way the system retains legitimacy. Satire, caricature and harsh parody remain protected; the red line is realistic deception meant to convince voters an event or statement actually happened.

The Shared Factual Floor Is Still the Stakes

Political spin, selective framing and hard attacks are normal. Democracy can survive them. It cannot survive competing versions of physical reality that look identical on a phone screen. Citizens must still be able to choose among visions for the country without first having to decide which videos are fabrications.

The new labels help on the official surface. They do little against foreign desks that ignore Israeli law, against domestic networks that stay deniable, or against the sheer speed at which synthetic media travels. The Comptroller’s findings and the INSS critique show the preparedness lag is structural, not a one-election glitch. AI simply made the cost of delay higher.

Comparable pressure appears elsewhere. The EU AI Act labeling requirements already in force show how large jurisdictions are racing the same technology. Inside Israel the campaign will also turn on ordinary coalition arithmetic; Democrats’ potential kingmaker role is one live variable among many. Polling already reflects public fear of internal divisions running as high as external threats. Synthetic media that deepens those divisions is not a side issue. It is the medium through which the fear can be engineered at scale.

  1. July 26, 2026, CEC temporary AI disclosure rules take effect for new campaign material
  2. Mid-July 2026, Likud AI Eisenkot-Abbas video draws national backlash
  3. July 26, 2026, Sohlberg orders Smotrich AI video deleted as misleading
  4. July 2026, State Comptroller special report published; TikTok election task force and guidelines announced
  5. October 27, 2026, Scheduled election day for the 26th Knesset

Fast complaint handling, meaningful penalties, equal application across camps, and a real national mechanism for foreign digital interference are still missing. Labels alone will not restore a common factual baseline. The October campaign will show whether Israel can close the remaining gaps before synthetic media sets the terms of the debate.

The rules are live. The harder test is everything they cannot yet reach.

Leave a Reply

Your email address will not be published. Required fields are marked *