On 15 October 2026, twelve days before Israel's 27 October election, the Central Elections Committee issued an interim order requiring the Yashar party, led by Gadi Eisenkot, to label or remove a digitally altered video. The video shows Prime Minister Benjamin Netanyahu on a phone call with Yahya Sinwar, congratulating him. It ends with a title card reading, in the Times of Israel's translation, "Netanyahu, you were warned, and you didn't prevent [it]." Likud petitioned under the Elections (Propaganda Methods) Law, and the hearing is ongoing.
This is the latest enforcement of the committee's July 2026 AI-disclosure rules. It is also a useful data point for a country with no comprehensive AI statute.
The strongest case for stricter rules
The case for going further than a label is serious. A synthetic clip of a prime minister speaking with a dead terrorist leader can spread faster than any correction. Many viewers will see the clip without the caption or the court order. A label on a video can also be cropped away when the video is reshared. Regulators who want takedowns, not disclosure, are responding to a real problem, and a campaign that lasts only weeks gives courts little time.
In a separate petition this campaign, the committee went further. JFeed reports that it ordered Bezalel Smotrich's party to remove an AI video showing Eisenkot shaking hands with Mahmoud Abbas, and to pay petitioners 10,000 shekels. The committee therefore has a removal remedy and uses it when it finds a video misleading.
What the rules actually require
The rules are narrow. They were published in the official gazette (Kovetz HaTakanot 12484) on 26 July 2026 under the Elections (Propaganda Methods) Law. The Hebrew text targets "deep" election propaganda: content that depicts a person saying or doing something they did not say or do, in a way that could pass as authentic. The disclosure must be on a plain background, in high-contrast colour, and in the language of the content. On images it must be at least 5% of the content's height or 10 points, whichever is greater. Video needs it throughout, and audio-only content needs it spoken. Pearl Cohen's summary adds that an attached notice creates a presumption of compliance, and that the rules add to other labelling and accuracy duties and do not replace them.
This is a transparency rule. It does not ask whether a message is true, flattering or fair, and it does not ban parody, satire or even a harsh fabrication, provided a viewer is told it is synthetic. Under the Yashar order, the party could keep an obviously partisan video up as long as it carries a label.
That is the pro-speech argument for the design. A political campaign is exactly the place where speech protection matters most. A label puts the burden on the publisher and leaves the viewer to judge. A removal order, as in the Smotrich case, should be reserved for content that is misleading in a way a label cannot repair.
Why this fits Israel's wider AI approach
Israel has deliberately avoided a horizontal AI law. The Ministry of Innovation, Science and Technology's December 2023 "Responsible Innovation" policy, prepared with the Ministry of Justice, relies on existing sectoral regulators, risk-based tools and soft law. The IAPP's analysis notes that its AI knowledge and coordination center, set up under Government Decision No. 173 of February 2023, has no enforcement powers, and that regulators keep their own. The same analysis cites a 2023 Israel Innovation Authority report putting high-tech at 18% of GDP and 50% of exports. Israel's AI-heavy tech sector is a core reason it has been wary of rules that could slow it down.
Election law is one place where a sector regulator already exists and already has enforcement tools. The committee did not need a new AI statute. It needed a rule defining a labelling duty under a power it already held, and it used it. That is the model the 2023 policy envisaged: empower the regulator that knows the domain, and keep the rule proportionate.
Where the model is weaker
There are limits, and honest analysis should state them. The rules are election-specific and temporary in nature, so they will not govern AI-generated defamation, fraud or non-consensual imagery outside the campaign. Enforcement is petition-driven: Likud had to file, and the committee can act only on what is put in front of it. That favours well-resourced parties, and it means the system depends on rivals policing each other. And a label on the original upload does not follow a clip through screenshots and re-uploads, as critics of disclosure rules point out.
There is also a risk that the committee's two remedies drift together. If labelling is treated as a mere formality and removal becomes the default for any provocative synthetic clip, the rules would turn from a disclosure regime into a content-approval regime. The Yashar order points the better way: it offers the party a choice between labelling and removing, and does not rule on the video's political message.
What to watch after 27 October
Three questions matter. First, whether the committee's final ruling in the Yashar case keeps the label-first approach, or adopts the stricter logic of the Smotrich decision. Second, whether the parties comply visibly, since a label that is technically present but unreadable would defeat the rules. Third, whether the Knesset and the Ministry of Justice draw on this experience for any permanent rules, or leave the temporary regime to lapse.
For other jurisdictions weighing deepfake rules, the Israeli approach offers a practical lesson. A narrow rule, enforced by a body with existing authority, specifying exactly what a label must look like, is easier to follow and to defend in court than a general ban on synthetic media. Israel has not solved AI misinformation, and the rules' main test is still ahead. But it has shown that serious enforcement is possible without a comprehensive AI law, and that is an argument for proportionate regulation elsewhere.