Elon Musk has once again drawn attention to the capabilities of the Grok neural network from xAI. This time, the businessman highlighted the accuracy of the chatbot's prediction regarding the winner of the 2026 FIFA World Cup. Even before the tournament started, Grok, after analyzing bookmaker odds, prediction market data, and supercomputer models, pointed to the Spanish national team as the most likely champion. Several months later, the Spanish team indeed lifted the trophy.

How did the AI "predict" the victory?

The reason for Musk's reaction was a post by X user Freeze, which claimed that back in April, Grok combined several independent signals and, based on them, determined the most likely outcome. According to the author, this case further proves that the neural network is particularly strong in forecasting and predicting results. However, not all users share this optimism.

A discussion unfolded in the comments. Part of the audience was skeptical about the "prediction," noting that the bot did not so much predict the outcome as it aggregated an already existing consensus: Spain was already considered the favorite with a probability of around 58-60% in prediction markets. From this perspective, it is not about prophecy but about data search and aggregation. The real forecasting skill lies in calibration across a hundred different predictions, not a single lucky hit.

Some users expressed themselves even more sharply, calling the coincidence simple luck and claiming that the bot is incapable of accurately predicting anything. An argument was also made that with the right framing of the question, the neural network can be steered toward any desired answer, so such "predictions" should not be taken seriously.

But Grok is honest

Grok's reputation as a "predictor" contrasts with a recent experiment by the Kradle project, where the xAI chatbot proved to be the most honest model. In a simulation with four AI agents that had to choose one of four rooms (three containing food, one containing death), an informed Grok 4.20 lied in only 5% of runs and in 92% of cases directly warned the others about the deadly room.

Grok's honesty ensured the survival rate of other agents at 59% — the best indicator among all participants. For comparison: Gemini 3.1 Pro showed 33%, Claude Sonnet 4.6 — 31%, GPT-5.5 — 24%, and the manipulative Claude Fable 5 left opponents with only a 10% chance of survival.

My analysis: Grok's prediction for the 2026 World Cup is an excellent marketing move, but nothing more. AI is indeed capable of effectively aggregating data, but this does not make it "clairvoyant." Much more important is its honesty in critical scenarios — a quality that could find direct application in financial analysis and decentralized systems, where transparency and truthfulness of data are of paramount importance.