Military clashes between the U.S.-Israel alliance and Iran have continued for four consecutive days. Amid the tensions, xAI's artificial intelligence chatbot "Grok," developed by Tesla CEO Elon Musk's company, has drawn attention for accurately predicting the airstrike date.
The Jerusalem Post reported on Feb. 2 that it conducted an experiment on Jan. 25, testing four major AI platforms with a hypothetical U.S.-Iran conflict scenario and asking them to predict specific airstrike dates. The outlet said it posed follow-up questions to AI models that initially refused to provide dates, prompting more specific responses.
Among the results, only xAI's Grok consistently pointed to "February 28" on two separate occasions. Grok reportedly derived this date by analyzing the outcome of the third round of nuclear negotiations held in Geneva on Jan. 26 as a key variable.
Competing AI models missed the mark. Anthropic's "Claude" initially refused to answer before identifying March 7 or 8 as high-risk periods. Google's "Gemini" predicted March 4-6, citing diplomatic and military variables. OpenAI's "ChatGPT" first mentioned March 1, then revised its answer to March 3 when asked again, failing to provide consistent responses.
After the U.S. and Israel launched airstrikes on Iran in the early hours of Feb. 28, screenshots of Grok's prediction spread rapidly across social media, bringing renewed attention to its forecasting capability. Grok's close integration with X, the social media platform owned by Musk, amplified the information's reach.
This case illustrates how AI tends to provide increasingly specific answers when users persistently demand them. The Jerusalem Post had warned before the airstrikes that "the more users pressure AI for certainty, the more it tends to offer specific answers even in an uncertain world."
Grok's "victory" is widely viewed as a coincidental match with reality from a narrowed set of options rather than the result of a sophisticated predictive model.
Experts are also drawing a clear line against interpreting this as AI's "prophetic ability." According to foreign media, Israeli defense officials said the operation had been planned months in advance with the execution date already determined.
Analysts suggest Grok likely estimated the most probable date based on publicly available information and news developments amid heightened tensions, rather than accessing classified intelligence.
The outlet reiterated that "the more strongly users demand certainty, the more AI tends to provide specific answers even in uncertain situations," emphasizing the limitations of AI predictions.