Algorithmic Refereeing at the 2026 World Cup: Compatibility with International Football Law and the Integrity of Competition

[Mohammad Saleh Anisi is an Assistant Professor at Science and Culture University, Faculty of law]

The 2026 FIFA World Cup was supposed to be remembered for its expanded format, its three host nations, and the spectacle of 104 matches across the continent. Instead, it may be remembered for a different kind of drama, one that unfolded not on the pitch but in the invisible architecture of algorithmic decision-making. In the 93rd minute of Iran versus Egypt on June 26, 2026, Shoja Khalilzadeh converted a scrambled rebound from a free-kick sequence in what appeared to be a historic winner — Iran’s first-ever path to the knockout stage seemingly secured. The bench stormed the field. Fans wept. Then, following a prolonged VAR review, the goal was disallowed. Khalilzadeh, the system determined, had been marginally beyond the second-to-last Egyptian defender at the moment of the flick-on, with the goalkeeper off his line; the relevant reference point was a field player, making the geometry of the call anything but obvious to the naked eye. No stadium screen explained the full chain of positional logic. No official articulated why the geometric relationship between three bodies in motion, frozen at a microsecond determined by software, constituted active offside interference. The referee signaled. The goal was gone. This is not merely a story about heartbreak. It is a case study in what happens when the decisional process becomes opaque, accountability becomes untraceable, and the legal framework governing referee authority has not kept pace with the technology it authorized. Under Law 5 of the Laws of the Game, the referee bears “full authority” to enforce the rules, but authority exercised through an algorithmic black box raises a question that sports law has not yet answered: whose authority, exactly, was that?

This moment crystallizes the central legal question this post addresses: how far can AI-assisted systems be expanded in football refereeing before they compromise the legitimacy of the human referee and the integrity of competition? The answer is not a question of technical capability — the Semi-Automated Offside Technology (SAOT) deployed at this World Cup is demonstrably more accurate than any human eye. It is a question of institutional design and legal compatibility. The expansion of AI in refereeing is compatible with international football law, but only conditionally: as long as these systems remain auxiliary tools, and final decision-making authority remains with the human referee. When that boundary erodes — not necessarily through formal rule change, but through the quiet accumulation of operational pressure — referee legitimacy weakens, and competition integrity becomes structurally compromised.

Before reaching the legal analysis, these two variables require precise definition, because vagueness here produces arguments that prove too much or too little. Referee legitimacy in sports law literature operates on two distinct planes: procedural legitimacy, which asks whether the decision-making process complied with established rules; and substantive legitimacy, which asks whether the decision-making authority is accepted as valid by the relevant stakeholders. A referee decision can be procedurally unimpeachable — fully consistent with the Laws of the Game — and yet substantively illegitimate if players, coaches, and spectators cannot identify who made the decision, cannot understand why, and cannot locate accountability. Competition integrity, meanwhile, is not synonymous with accuracy. FIFA defines it broadly as the assurance that results are “achieved through the equal, transparent and trustworthy application of the rules.” Accuracy is a necessary condition of this assurance, but not a sufficient one. A system that produces accurate outcomes through an opaque process — one whose decisional logic cannot be interrogated, whose training data cannot be audited, and whose outputs cannot be attributed to a responsible human agent — does not satisfy the integrity standard even if it never makes a factual error.

The foundational legal text is Law 5 of the Laws of the Game, as maintained by the International Football Association Board (IFAB). In its 2025/26 edition, Law 5 vests the referee with “full authority to enforce the Laws of the Game during the match,” making the referee’s decisions on “matters relating to play” final. Crucially, the same provision permits competitions to use technology “to assist referees in making or changing decisions,” but the grammatical architecture of this permission is normatively decisive: technology assists the referee — it does not substitute for her. The distinction between “assisting the referee in making a decision” and “making the decision that the referee then announces” is not merely semantic. It marks the boundary between a system that enhances human judgment and a system that displaces it. IFAB’s drafting choice reflects an institutional commitment to what might be called the human-authority model of refereeing — a model in which the referee is not an administrator of algorithmic outputs but a sovereign decision-maker who may, or may not, consult technological input. The legal significance of this framing cannot be overstated: it is the textual foundation upon which any assessment of AI’s permissible scope in refereeing must rest.

The VAR protocol, introduced at Russia 2018 and codified in Law 6 of the Laws of the Game, represents the most elaborate pre-2026 instantiation of the auxiliary tool model, and its design reveals how carefully IFAB has historically policed the boundary. The protocol confines VAR intervention to four categories of “match-changing situations”: goal/no goal, penalty/no penalty, direct red card, and mistaken identity. More revealing than the scope limitation is the intervention threshold: VAR may only flag a “clear and obvious error” or a “serious missed incident.” This threshold reflects a deliberate institutional choice to preserve referee discretion across the vast space of contested and borderline calls, to ensure that technology supplements the human referee’s judgment rather than replacing it in precisely those situations where reasonable people disagree. Even within these narrow categories, Law 6 is unambiguous: “the final decision is always made by the referee.” A referee who exits the review area and maintains her original decision has not committed a procedural error — she has exercised the authority Law 5 grants her. Competition integrity, in this model, is served precisely because there is a human being who saw, judged, and can be held responsible.

The 2026 tournament’s upgrade to SAOT introduces a structural modification that deserves more legal attention than it has received. In Qatar 2022, the SAOT alert traveled first to a human VAR official who then relayed it to the on-field referee. In 2026, the alert is transmitted directly from the algorithm to the on-field referee’s earpiece, bypassing the human intermediary. IFAB has constrained this direct channel to “positional offside only” cases where a body part is measurably beyond the second-to-last defender, and has explicitly excluded interference-in-play determinations from the system’s remit. On paper, the human-authority model is preserved. In practice, the picture is more complicated. When an algorithm with ten-centimeter precision sends an alert simultaneously with a three-dimensional avatar appearing on stadium screens for tens of thousands of spectators and a global television audience, the referee who overrides that alert faces an accountability burden that her predecessors never encountered. The formal authority vested by Law 5 remains intact; the real authority to exercise independent judgment is structurally diminished. This is the gap between de jure and de facto decision-making power, and it is precisely the gap that Iran’s disallowed goal in the 93rd minute made visible.

This structural shift has normative implications that extend beyond the text of IFAB’s Laws. In international sports law, the concept closest to opinio juris — the settled normative expectation that a practice is legally required, not merely operationally convenient — helps identify when a departure from established practice constitutes a violation rather than merely an innovation. The acceptance of the auxiliary-tool model across decades of IFAB rulemaking, VAR implementation, and CAS doctrine has generated just such a normative expectation: technology in football refereeing exists to support human judgment, not to supplant it. A 2025 empirical study of German football officials found that the perceived success of refereeing technology depended not on computational accuracy but on the preservation of referee autonomy relative to the system. When that autonomy erodes in practice — as the SAOT’s direct alert channel risks producing — the normative expectation is breached, regardless of whether a formal rule has been amended. The consequence is not just a procedural gap but a substantive legitimacy deficit: stakeholders no longer know whose judgment they are accepting when they accept a decision.

The Court of Arbitration for Sport’s field-of-play doctrine, developed across a line of decisions from CAS 2010/A/2090 through the Paris 2024 Olympic cases, establishes that refereeing decisions are generally non-reviewable in appellate proceedings absent proof of “bad faith, arbitrariness, or a legal error.” This doctrine rests on a coherent institutional rationale: appellate bodies are not sport specialists; they were not present at the moment of decision; and systematic review of technical calls would undermine referee authority and the finality of results. This rationale, however, presupposes a specific ontology of refereeing: decisions emanate from an identifiable human agent who perceived, evaluated, and judged. The concepts of “bad faith” and “arbitrariness” are intelligible only in relation to a subject with a will capable of deviation. An algorithm has no will, no good or bad faith, and cannot be arbitrary in the legal sense — it processes data according to its training and architecture. If the real decision in the Iran case was generated by a system whose training data, decisional logic, and potential systematic biases are not publicly auditable, then the field-of-play doctrine creates what AI ethics scholars call a “responsibility gap”: a situation in which no human agent can be held accountable for the decision, but the algorithmic system cannot be held accountable either. This gap is not merely theoretical. It is the structural condition that the 2026 World Cup has created and that existing sports law has not closed.

The competition integrity implications follow directly. If the process by which a match-deciding call is made cannot be explained to the players it affects, attributed to an accountable human agent, or subjected to meaningful review, then the result — however technically accurate — does not satisfy the transparency and trustworthiness requirements that underpin competition integrity. Iran’s players in Seattle could accept a referee’s judgment that their player was offside; referees make calls, and that is the nature of the game. What they cannot meaningfully accept or contest is a decision whose justification exists only in the parametric space of a machine-learning model that neither FIFA, IFAB, nor any public authority has disclosed in auditable form. Technical accuracy, divorced from process transparency and human accountability, does not produce legitimate outcomes. It produces the appearance of legitimate outcomes, which is something fundamentally different.

The conclusion this analysis compels is clear. The expansion of AI-assisted systems in 2026 World Cup refereeing is legally compatible with international football law, but only so long as three conditions hold simultaneously. First, the referee must retain not merely formal but real decisional authority: the operational and institutional design must preserve her genuine capacity to override algorithmic alerts without facing accountability consequences that effectively render override impossible. Second, every match-deciding call must be attributable to an identified, accountable human agent — not because the algorithm necessarily erred, but because the legal architecture of sports adjudication, from IFAB’s Laws to CAS doctrine, is built around human agency and cannot function otherwise. Third, the decisional process must be sufficiently transparent for stakeholders to evaluate — not merely to observe outputs on stadium screens, but to assess whether the underlying system is free from systematic bias and applied consistently across all matches. If all three conditions hold, AI makes refereeing better. If any one fails, football has not achieved better officiating. It has achieved the appearance of better officiating, and that appearance, as the 93rd minute in Seattle demonstrated, can alter World Cup fates and leave an entire nation unable to say, with any confidence, who exactly made the decision that changed everything.

Photo by Fauzan Saari on Unsplash