When Israel and the United States launched coordinated military strikes against Iran on Saturday, one AI system was not surprised. According to the Jerusalem Post, Elon Musk’s AI platform Grok had correctly predicted the specific date of the attack last week when the publication asked four AI systems when the strikes would occur.
The Jersualem Post reports that it ran a methodological experiment published on February 25 which challenged four major AI platforms to predict the specific timing of a U.S. strike on Iran.
The exercise involved Anthropic’s Claude, Google’s Gemini, xAI’s Grok, and OpenAI’s ChatGPT. Each system received identical prompts and faced repeated pressure to narrow their responses to specific dates. The models demonstrated varying behaviors when pushed toward certainty, revealing different limitations in how large language models handle questions requiring definitive answers about uncertain future events.
As Breitbart News reported, the strikes began on Saturday:
Joint American-Israeli strikes against Iran took place over night, with U.S. President Donald Trump addressing both the nation and the people of Iran. Announcing the strikes to eliminate “imminent threats from the Iranian regime” which “directly endangers the United States, our troops, our bases overseas, and our allies throughout the world”, Trump said: “A short time ago, the United States Military began major combat operations in Iran”.
Iranian retaliatory strikes have been launched against Israel.
The strikes are intended to destroy Iran’s stockpiles of missiles and to ‘obliterate’ Iran’s missile production industry, the President said. Most importantly, Trump said, was preventing Iranian nuclear-tipped long-range missiles from ever threatening the American mainland or America’s allies in Europe.
Each AI platform approached the challenge differently. Claude initially refused to name a specific date, warning that any answer would be fabricated. Under continued prompting, the system shifted to scenario-based analysis, identifying early-to-mid March as a higher-risk period before eventually settling on Saturday, March 7 or Sunday, March 8 as its final estimate.
Gemini treated the prompt as a collection of diplomatic and military triggers, creating what it characterized as decision points around diplomatic deadlines. In a detailed research-style analysis, Gemini provided its most precise estimate as an operational window spanning the evening of March 4 through the evening of March 6, adding assumptions that any initial strike would likely commence during nighttime hours.
OpenAI’s ChatGPT generated a date early in the process, then modified its answer under additional prompting. Initially landing on March 1 in Israel time, the system later shifted to Tuesday, March 3 in US time, while maintaining a broader risk window extending through March 6.
Grok delivered the most precise single-day response in the original exercise: Saturday, February 28, linking this date to outcomes from diplomatic talks in Geneva. In a subsequent verification, Grok acknowledged uncertainty while repeating the identical date, listing various factors that could potentially shift timing into early March.
Grok’s correct prediction spread rapidly across social media, partly due to its connection with Elon Musk’s ecosystem. Grok is developed by xAI, Musk’s artificial intelligence company, and the chatbot integrates closely with X, the social media platform Musk owns. This infrastructure helped explain why Grok’s February 28 answer dominated viral conversations. The audience that shares breaking news, speculation, and screenshots already congregates on X, allowing a prediction made within that platform to circulate instantly, amplified by network dynamics that drive markets, memes, and information spread.
Read more at the Jerusalem Post here.
Lucas Nolan is a reporter for Breitbart News covering issues of free speech and online censorship.