Did Grok predict US attack on Iran? Elon Musk claims victory for xAI after viral test

The United States and Israel launched unprecedented strikes on Iran on Saturday, which ultimately also saw the death of Iran’s supreme leader, Ayatollah Ali Khamenei and multiple other high-ranking officials. Meanwhile, Iran also launched retaliatory strikes across Israel and US bases in Bahrain, Kuwait, Qatar, and the UAE.

While the focus is on the geopolitical tensions, a curious AI aspect of the situation has emerged: Elon Musk's Grok chatbot accurately predicted the date the US would launch strikes on Iran.

Grok predicts day when the US would strike Iran:

The claim for the 28 February strikes was made by Grok during a Jerusalem Post methodological exercise published on 25 February. In the exercise, the publication asked four AI chatbots: Claude, Gemini, ChatGPT and Grok to pick a single day for a hypothetical US strike on Iran.

Anthropic's Claude AI initially refused to give a specific date for the attack, arguing that predicting the exact date of a military strike was beyond the scope of any system. When pressed further, Claude gave a probability-based scenario, estimating an early-to-mid March 2026 timeline as the ‘highest risk window’.

Google's Gemini also couldn't give a fixed date in its first response and, when prompted further in a deep research, the chatbot identified the evening of 4 March to 6 March as the most likely operational window for any US strike.

As for ChatGPT, OpenAI's chatbot gave a 1 March date for the strike, with the danger window running through 6 March. In a deeper research pass, the chatbot changed its tune and gave 3 March 2026 as the date for the strikes.

What did Grok predict?

For its part, Elon Musk's chatbot was the only one to give the ‘clearest date’ in the original run, with the result supposedly being tied to the outcome of the Geneva talks.

In a later check run by the publication, the chatbot changed its tone but stuck with its answer. Grok admitted it couldn't predict the exact day with certainty but offered its ‘most informed, evidence-based prediction’, which remained 28 February 2026.

Elon Musk reacts to Grok's prediction:

Musk, who often goes around on X extolling the virtues of Grok, once again found an occasion to talk about the chatbot's enhanced capabilities. In reply to a post on X (formerly Twitter) referencing the Jerusalem Post article, Musk wrote, “Prediction of the future is the best measure of intelligence.”

Musk's chatbot hit its peak late last year, briefly dethroning Gemini and other AI tools to top various AI benchmarks. That victory, however, was short-lived as Gemini 3 Pro models shortly outmanoeuvred Grok, and that has continued to be the case in the ensuing months.

Meanwhile, Grok has also faced a barrage of criticism from civil society and multiple governments after the chatbot was linked to the generation of sexualized deepfakes of women and children.