How the 2026 World Cup Exposes AI Regulation Gaps

Croatia’s disallowed equalizer against Portugal is the cleanest way to see the governance problem inside the 2026 World Cup’s AI-assisted officiating debate. The controversy was not that a referee missed something obvious. It was that a connected ball detected a touch that human observers could not see, and that machine-readable event helped decide whether a goal existed in the match record at all.[1]

That distinction matters. A traditional officiating error is visible enough to argue over: a flag raised too early, a foul judged harshly, a replay angle that appears inconclusive. The Croatia incident moved the decisive fact somewhere else. The relevant touch was not available to the players, the crowd, or the referee’s unaided senses. It existed inside a technical pipeline, then emerged as an authoritative input into a binding competitive decision.

Calling that system “assistive” does not settle the question. Assistance can still become authority when the human actor has no practical basis to contradict the machine. The referee remains visible, but the basis of the decision has shifted away from ordinary perception and into an infrastructure that only a narrow group can inspect.

The System Is Precise Enough to Create a Governance Problem

Semi-automated offside technology is not simply VAR with a sharper camera. FIFA describes a system built from 12 dedicated tracking cameras installed under the stadium roof, following the ball and up to 29 data points on each player, 50 times per second, to calculate player positions relevant to offside decisions.[2]

The connected ball adds another layer. FIFA’s documentation says the official match ball contains an inertial measurement unit sensor positioned at the center of the ball, sending ball data to the video operation room 500 times per second. That feed is used to identify the kick point with far greater temporal precision than a broadcast replay can supply.[2]

The elegance of that engineering is not the problem. It is exactly why the governance issue is serious. The system can detect events too fine-grained for normal spectatorship, too fast for ordinary human review, and too technical for a referee to verify independently during a match. Once those events are folded into VAR review, they become operationally decisive.

FIFA’s model still places a human video match official in the loop. The system generates an automated offside alert, video officials validate the proposed decision, and a 3D animation can be produced for stadium and broadcast audiences after the decision has been confirmed.[2] That sequence gives the process a human checkpoint, but it does not give the affected player, team, or public a meaningful way to interrogate the underlying sensor event.

Human Oversight Is Doing More Work Than the Phrase Can Bear

In a well-governed automated decision system, “human oversight” has to mean more than a person receiving the output. The human must understand enough about the system, its limits, and the evidentiary basis for the output to intervene when intervention is warranted. In SAOT-assisted offside review, the referee can review the resulting visualization and the VAR recommendation, but the decisive data stream is not something the referee can reproduce with eyesight or ordinary match evidence.

The Croatia case exposes the asymmetry. A player can be told that the ball sensor detected a touch. A coach can dispute the sporting consequence. A broadcaster can show an animation. None of those acts amounts to an independent explanation of how the system distinguished a legally relevant touch from noise, how confidence was assessed, or what failure modes were considered in that specific decision.

The visible official also absorbs institutional risk. When a disputed decision is announced, the referee remains the public face of the outcome. Yet the referee did not design the sensor, validate the data pipeline, set the confidence thresholds, certify the camera calibration process, or choose the disclosure policy. The official is accountable in the match narrative for a system whose most important components sit outside the official’s control.

That is why the fairness question cannot be reduced to accuracy. Accuracy is indispensable; a less accurate system would be worse. But accuracy alone does not answer who validated the tool before deployment, who can examine the evidence after a disputed decision, who has authority to challenge it, and who bears responsibility if the technology malfunctions.

The EU AI Act Is a Useful Counterfactual, Not a Direct Cause of Action

The EU AI Act does not turn the 2026 World Cup officiating system into an obvious legal violation. The tournament is not being governed as a laboratory for EU AI Act enforcement, and sports officiating does not neatly drop into the Act’s listed high-risk categories. The comparison is still useful because the Act shows what modern AI governance increasingly treats as essential when automated systems shape consequential decisions.

For readers tracking the Act’s classification structure, the relevant question is the one usually asked in a high-risk AI Act analysis: what happens when a system’s output materially shapes a decision that affects someone’s rights, opportunities, or status? The Act’s broader risk structure, summarized in this EU AI Act risk categories glossary, is built around the idea that some AI systems require obligations before and during deployment, not merely explanations after public controversy.

If a comparable system were treated as high-risk under Annex III logic, the regulatory architecture would look different. The EU AI Act requires high-risk AI systems to meet obligations including transparency and instructions for use under Article 13, human oversight under Article 14, and conformity assessment procedures under Article 43.[3] The point is not that FIFA has breached those provisions. The point is that sports governance lacks an equivalent architecture for a system with real competitive finality.

The missing piece is not a courtroom on the touchline. Football cannot pause indefinitely while counsel cross-examines a sensor log. But procedural legitimacy does not require turning a match into litigation. It requires that the institution using the system can answer basic questions before the tournament begins and after a disputed decision occurs.

Sports Governance Has Legal Risk Categories, but Not a Full Architecture

A legal analysis by O’Connors identifies three areas where semi-automated offside technology creates pressure: data privacy, liability allocation, and transparency or fair-play obligations. On privacy, the firm treats player biometric data as personal data under UK GDPR analysis; on liability, it asks who is responsible if the technology malfunctions; on fair play, it links transparency to public trust in the competition.[4]

That analysis is useful because it separates issues that are often collapsed into a single complaint about technology. Privacy asks what data is collected and under which legal basis. Liability asks who pays, apologizes, or corrects the record when the system fails. Transparency asks whether affected parties can understand enough about the system to accept its authority as legitimate.

The jurisdictional limits matter. A UK GDPR framing does not automatically govern every processing activity in a tournament hosted across the United States, Canada, and Mexico, with players and teams connected to many legal regimes. The privacy question is therefore not solved by naming one statute. It is sharpened by the fact that the tournament’s technical system operates across a fragmented legal environment.

The broader ethics literature has already noticed the same pressure. A systematic scoping review of AI ethics in sport examined 25 studies across 9 databases and found that 22 of the 25 identified privacy and data ethics as a primary concern.[5] That does not prove SAOT is unlawful or unethical. It does show that the data-governance problem is documented, not invented after one disputed match.

Operational Stress Makes the Oversight Gap Harder to Ignore

VAR already places officials inside a demanding operational environment: real-time pressure, broadcast scrutiny, competing replay angles, and immediate institutional consequences. SAOT can reduce some factual uncertainty, especially around offside geometry and kick-point timing, but it also adds a dependency on technical systems that officials may not fully understand.

The experience gap is not evenly distributed. According to former FIFA referee Armando Archundia, only 20% of countries participating in the 2026 World Cup had VAR in their domestic competitions, a figure that should be treated as directional because it is not independently audited against a FIFA dataset in the available materials.[1]

A Northeastern University analysis published on July 11, 2026, reported more than 100 VAR interventions by the end of the Round of 16, with approximately 0.5 decisions overturned per match, higher than in 2022. Because that analysis appeared before the tournament concluded, it is best read as a mid-tournament snapshot rather than a final tournament-wide measurement.[6]

These figures do not prove that SAOT is failing. They show something narrower and more important for governance: a technically complex decision system is being used in a setting where institutional readiness, referee experience, public explanation, and competitive stakes are unevenly aligned.

The Appeal Problem Is Structural

Football has never offered a general right to relitigate every refereeing decision. That is part of the sport’s structure. But automated evidence changes what the absence of appeal means. When a human referee makes a judgment call, the lack of appeal preserves the authority of on-field adjudication. When an opaque technical system supplies the decisive fact, the lack of appeal protects the system from scrutiny.

A meaningful challenge process would not need to overturn results casually. It could be narrow: preserving system logs, identifying which data stream supplied the decisive fact, documenting whether the relevant cameras and sensors were functioning within approved parameters, and specifying who reviewed the anomaly if one was alleged. Those are governance questions before they are fan grievances.

The current public-facing model relies heavily on trust in FIFA, the technology providers, and the match officials. Trust may be deserved in many decisions. It is still not the same as transparency, and it is not a substitute for redress when the system’s output is both opaque and practically final.

A Sports Case With Consequences Beyond Sport

A preprint analysis has characterized this kind of system as “illegitimate algorithmic governance,” a useful phrase so long as it is handled carefully: the available research materials support the general concern that automated systems can influence binding competitive outcomes without procedural safeguards, but the full preprint argument is not available here for detailed verification.[7]

The phrase captures why the World Cup example travels. Real-time automated decision systems are spreading into settings where delay is costly and institutional actors prefer quick outputs: security screening, workplace monitoring, insurance triage, education integrity systems, and emergency operations. In each setting, the same institutional temptation appears. Keep a human in the formal decision chain, describe the technology as assistive, and avoid building the heavier machinery of explanation, oversight, and contestation.

The 2026 tournament has also produced non-AI governance controversies, including the Balogun-Trump red card dispute. That episode belongs to a different category: institutional independence and political pressure, not automated decision regulation. It is still a reminder that legitimacy in sport depends on more than correct outcomes. It depends on whether the decision-making institution can show why its authority deserves to be accepted.

SAOT may be accurate. It may improve many offside decisions. It may spare referees from some impossible calls. None of that answers the harder question exposed by Croatia’s erased equalizer: when the decisive fact comes from a sensor no one in the stadium can independently perceive, what procedure makes that fact governable?

The 2026 World Cup does not prove that semi-automated offside technology is unlawful or useless. It does show how institutions can deploy AI-like decision systems with high practical stakes while avoiding the transparency, oversight, and redress architecture that modern AI regulation increasingly treats as essential.

References

1. The Problem With VAR at the 2026 World Cup Isn't the Technology—It's Who Interprets It, WIRED
2. Semi-automated offside technology, FIFA
3. Regulation (EU) 2024/1689, EUR-Lex, 2024
4. The legal implications of semi-automated offside technology in professional football, O'Connors
5. Artificial intelligence ethics in sport: a systematic scoping review, PMC
6. VAR at the World Cup: Why video review is causing so much controversy, Northeastern University, July 11, 2026
7. Illegitimate Algorithmic Governance in Sports, Preprints.org