One in Eight Election Posts in Israel Is AI-Made. The Same Cheap Fakes Are Already Aimed at Your Money
Israel’s Oct. 27 vote is being called the first election of the AI age, policed by rules written for radio. The tool that fakes a prime minister asleep also fakes a grandchild in trouble — here’s the household defense.
By Sean Anees Saifi · Capital Wealth · Published Monday, September 21, 2026 · Source: The Wall Street Journal, Monday, September 21, 2026 edition, whose market figures are the Friday, September 18 close
Key Points
Israel votes Oct. 27, its first election since the Hamas-led attacks of Oct. 7, 2023 and the Gaza war; AI researchers and developers call it the first general election of the AI age.
Ze Be AI, a research project at the Open University of Israel, finds roughly 1 in 8 election-related social posts in recent months was created or altered with AI, and that Prime Minister Netanyahu’s Likud party accounts for 46% of those posts.
Politicians, the prime minister included, have posted AI videos showing a rival asleep as attack warnings flash or judges dragged through the streets; each carries a small AI label and a Hebrew disclaimer that it was made digitally.
A live dashboard from the Brandeis Institute, an Israeli think tank, flagged 166 undisclosed AI posts last month; about 98% of what it flags never reaches the election committee, which has ordered only a handful of videos removed.
Israel’s election rules date from the 1950s and cover radio and print, patched temporarily each cycle; the Central Elections Committee has run dozens of drills with police and the Shin Bet, including a fake election-day claim that a city’s polling stations had closed.
1 in 8
election posts made or altered with AI, per Ze Be AI
46%
share of those AI posts traced to Likud accounts
166
undisclosed AI posts flagged by one dashboard last month
98%
flagged posts that never reach the election committee
Cheap AI video has turned campaigning into a contest of quantity, one Israeli AI executive told the paper. The same math works for scams.
In one line: The technology that fakes a prime minister asleep in bed is the one that fakes a grandchild’s voice on the phone, and the first regulator in your kitchen is a code word.
The prime minister is asleep in bed while warnings of a Hamas attack flash across the screen. Senior judges are dragged through the streets. The mayor of New York throws a party to celebrate Prime Minister Benjamin Netanyahu’s downfall. Each clip carries a small AI label and a Hebrew disclaimer saying it was made digitally and never actually happened — and each was posted by some of Israel’s most senior politicians, the prime minister among them.
Anat Peled reports from Tel Aviv that Israel’s Oct. 27 vote — the first since the Hamas-led attacks of Oct. 7, 2023 and the war in Gaza, and widely seen as a referendum on Prime Minister Benjamin Netanyahu — is being cast as the first general election of the AI age. Ze Be AI, a research project at the Open University of Israel, finds roughly one in eight election-related social posts in recent months was created or altered with AI, and that Netanyahu’s Likud party accounts for 46% of them. The rules meant to police this date from the 1950s and were written for radio and newsprint; they get patched temporarily each cycle, and researchers say the patches can’t keep pace.
The gap shows in the numbers. A live dashboard run by the Brandeis Institute, an Israeli think tank, flagged 166 undisclosed AI posts last month, and about 98% of what it flags never reaches the Central Elections Committee, which has ordered only a handful of videos taken down. The committee has run dozens of drills with police and the Shin Bet, including one imagining election-day posts falsely claiming a city’s polling stations had closed; its acting director general, Dean Livne, has overseen nine elections. Yonatan Zinder, who runs the Israeli startup Blueye AI, says cheap AI video has turned campaigning into a contest of quantity rather than quality, and Tehilla Shwartz Altshuler of the Israel Democracy Institute calls it “the first real AI elections in the world.”
Our read
We’ve got no view on Israeli politics, and the lesson that travels isn’t about elections. It’s how cheap it has become to fabricate a convincing person, and the same tools are already at work on households here. The grandchild who calls in tears from a jail cell, voice cloned from a snippet of video. The boss whose voice asks accounting for an urgent wire. The video call from a bank that looks exactly like your bank. A campaign uses these to move votes; a fraudster uses them to move money, and the fraudster has the same quantity advantage Zinder describes — it costs nothing to try you a hundred times.
So run the household the way the election committee ran its drills: before the day. Agree on a family code word that any real emergency call has to include. Never act on a call or a video; hang up and call back on a number you already have. Make it a rule that no money moves on urgency alone, because urgency is the whole product. And use the safeguard that brokerage firms must ask for under a FINRA rule, and some banks offer too: name a trusted contact on every account that allows one, someone the firm can reach when a request looks wrong. It takes ten minutes, and it’s a check that can still work after the fake has already fooled you.
What It Means For Your Portfolio
Hold — code word, call back, no money on urgency, name a trusted contact
The technology that fakes a prime minister asleep in bed also fakes a grandchild in tears and a boss asking for a wire. Israel’s election committee ran its drills before election day; households should too.
General planning principles, not advice for anyone in particular. Cheap AI video and voice cloning are consumer-protection risks, not just political ones. Set a family code word for any emergency call, call back on a number you already hold, and adopt a house rule that no money moves on urgency alone.
Name a trusted contact on every brokerage account — firms must ask for one under a FINRA rule — and on bank accounts that offer it. The contact can’t move your money; the firm can reach them when a request looks off. It’s a ten-minute form, and it’s a layer that can still work after the fake has already fooled you.