Point of Focus

• Coinbase published an AI-generated World Cup result before the match started.
• The error raises concerns over Coinbase’s prediction market strategy.
• CEO Brian Armstrong confirmed an investigation is underway.

Coinbase is facing criticism after an AI-generated notification incorrectly declared the outcome of a FIFA World Cup match before kickoff, raising fresh concerns about the reliability of artificial intelligence in financial products and prediction markets.

The incident prompted a public response from CEO Brian Armstrong, who said the company was investigating the issue, while other executives confirmed that the faulty alert had been removed and safeguards strengthened to prevent similar errors.

AI alert published a match result before the game started

The controversy began after Coinbase users received an AI-generated prediction market update claiming Norway had defeated Brazil 3-2, with striker Erling Haaland scoring twice. The notification presented the fabricated result as a completed event even though the match had not yet been played.

At the time the alert was sent, Coinbase’s own prediction market page showed that the knockout-stage fixture at MetLife Stadium in New Jersey was under a weather delay. No official result existed, making the notification factually incorrect.

this is what happens when a crypto company uses AI to generate sports prediction markets @coinbase is hallucinating results
for a World Cup game that hasn’t even been played yet and sending factually incorrect notifications to its millions of users as “breaking news”… pic.twitter.com/coD8xY2O0S

— jay (@jay_drainjr) July 5, 2026

Screenshots of the alert quickly spread across social media, with users accusing the exchange of allowing an AI system to “hallucinate” news and distribute it to millions of customers.

Coinbase CEO Brian Armstrong acknowledged the reports shortly after they surfaced.

“Taking a look with the team – thx for reporting it,” Armstrong wrote on X.

Later, Coinbase Chief Product Officer Max Branzburg confirmed that the company had removed the inaccurate story and introduced changes designed to reduce similar mistakes in the future.

“We fixed the incorrect story and made some updates to avoid these types of inaccuracies in the future,” Branzburg said, adding that while AI can deliver “24/7 insights for trading,” the system still requires further refinement.

Incident raises questions about Coinbase’s AI strategy

The mistake comes at an awkward time for Coinbase, which has increasingly integrated artificial intelligence across its products and internal operations.

Armstrong has repeatedly emphasized AI as a core productivity tool inside the company. Last year, he revealed that roughly 40% of Coinbase’s code was being generated with AI assistance, with ambitions to push that figure beyond 50%.

~40% of daily code written at Coinbase is AI-generated. I want to get it to >50% by October.

Obviously it needs to be reviewed and understood, and not all areas of the business can use AI-generated code. But we should be using it responsibly as much as we possibly can. pic.twitter.com/Nmnsdxgosp

— Brian Armstrong (@brian_armstrong) September 3, 2025

The company has also expanded AI-powered consumer features, including automated market insights and prediction market notifications intended to help users stay informed about real-time events.

The latest incident, however, highlights one of the biggest challenges facing AI-powered financial services: ensuring factual accuracy before distributing information that could influence trading decisions.

Prediction markets face renewed scrutiny

The incident also casts renewed attention on Coinbase’s growing ambitions in prediction markets.

The exchange recently expanded prediction market offerings across the US through its “Everything Exchange” initiative, with early liquidity supported by Kalshi.

Armstrong has described prediction markets as one of the most effective mechanisms for discovering truth, arguing that markets with financial incentives often produce more reliable information than traditional media.

That philosophy is now being questioned after an AI system attached to Coinbase‘s platform generated a fictional sporting result before the event had even begun.

The company has dealt with notification-related issues before. Earlier this year, Armstrong addressed a separate bug that caused unwanted trading alerts to be sent to users, saying Coinbase preferred fixing targeting errors over taking an overly restrictive approach to what customers could trade.