The largest American cryptocurrency exchange, Coinbase, found itself at the center of an awkward incident involving the use of artificial intelligence. A user under the nickname jay_drainjr reported receiving a push notification from the platform with the final score of a World Cup playoff match between Norway and Brazil. The problem is that the game had not yet started at that time.
According to the false alert, Norway won with a score of 3:2, and forward Erling Haaland scored a brace. In his post, the user called the incident "dangerous and irresponsible," emphasizing that it is a clear example of AI hallucinations, where the algorithm presents fabricated data as real news.
Management Response and Error Correction
Coinbase CEO Brian Armstrong quickly responded, promising to investigate the situation with his team. The very next day, the head of consumer and business products, Max Branzburg, confirmed that the error had been fixed and systems updated to prevent similar inaccuracies from recurring.
Notably, the actual match result differed from the AI's prediction: Norway did win, but with a score of 2:1. Haaland, as the algorithm "claimed," scored two goals. Branzburg humorously commented on this coincidence: "The AI might have known something we didn't."
Technical Details and Context
The false notification was generated in the "Trends" section, where AI automatically creates news summaries based on data from the prediction platform Kalshi. Recall that Coinbase recently integrated Kalshi as part of its "Everything Exchange" strategy — the concept of turning the exchange into a universal trading platform.
This incident occurs against the backdrop of active AI implementation in Coinbase's business processes. In May, the company cut 14% of its staff, citing the need to optimize expenses and focus on AI technologies. In June, the Coinbase for Agents service was launched, allowing users to connect digital assistants to their accounts for automated trading.
My expert assessment: This case is a classic example of the danger of uncontrolled use of generative AI models in financial products. When it comes to prediction markets and real money, even a single algorithm error can undermine the trust of millions of users. Coinbase was lucky that the false result partially matched reality; otherwise, the reputational consequences could have been far more severe. The market must learn this lesson: AI is a powerful tool, but without strict verification filters, it can become a source of misinformation.