xAI’s Grok chatbot has made headlines after correctly predicting the date of US and Israeli military strikes against Iran three days before they occurred, as revealed in a test by the Jerusalem Post published on February 25.
The Jerusalem Post conducted a test involving four AI models – Anthropic’s Claude, Google’s Gemini, xAI’s Grok, and OpenAI’s ChatGPT – to determine when the strikes would happen. Out of these, only Grok accurately identified the correct date of February 28, predicting “a limited US strike on February 28, 2026.” In contrast, the other models suggested dates in early March, with Claude settling on March 7 or 8, Gemini projecting March 4 to March 6, and ChatGPT revising its forecast to March 3.
The US and Israel indeed launched coordinated attacks on February 28, as Grok had predicted. Israel’s operation was codenamed “Roaring Lion,” while the US operation was dubbed “Operation Epic Fury.” The strikes were announced by President Donald Trump in a video address, with reports of explosions in Tehran, Isfahan, Qom, Karaj, and Kermanshah. According to the Associated Press and Reuters, Iran’s Supreme Leader Ayatollah Ali Khamenei was killed in the strikes.
In response to the attacks, Iran launched retaliatory strikes against Israel and US facilities in Bahrain, the United Arab Emirates, and Qatar. Elon Musk commented on Grok’s prediction on X, stating, “Prediction of the future is the best measure of intelligence.”
The Jerusalem Post characterized the exercise as a stress test rather than a forecasting service, noting that Grok’s prediction drew on publicly available signals, including Geneva diplomatic talks and Trump’s stated deadline from February 19. At the time, Reuters reported that a senior US official had suggested mid-March before all forces would be in place.
The Jerusalem Post concluded that the AI models responded based on the information available on the internet at the time, stating that “the robots answered when the internet asked for a date.” The test results were published on February 25, and Grok’s prediction quickly circulated on X through screenshots. The newspaper observed that pushing the AI models harder resulted in more specific answers, even though real-world clarity did not improve. The accuracy of Grok’s prediction may reflect either its analytical capability or coincidence in an exercise designed to test the models’ limits.
