AI Coaching in Esports: Exclusive Deals, Copycat Risk and the Grey Zone of the Rulebook

AI Coaching in Esports: Exclusive Deals, Copycat Risk and the Grey Zone of the Rulebook

Opening

In the interview with Jack Williams about iTero, two section headings sit side by side. The first covers an exclusive partnership with GIANTX and the likelihood of the tool being copied. The second covers AI-assisted cheating. One belongs to a commercial frame. One belongs to an integrity frame. Between them lies something esports has never named: ownership of a competitive advantage.

I read the summary several times looking for a timestamp to hold on to. No patch. No server version. No win rates, no rosters, no brackets. Every fact heavy enough to dissect sits outside the page. What remains is a story about contracts, about software licensing, and about a legal concept that was never finished.

Someone whose job is rebuilding referee decisions millisecond by millisecond tends to skip this kind of piece, because there is no slow-motion replay to scrub through. Yet precisely because there is no slow-motion replay, it is the clearest lesson in how a legal vacuum gets filled by contracts instead of regulation, if the league operator does not choose to fill it first.

One small detail deserves a pause. The piece mentions Natus Vincere and the Aegis of Champions at Gamescom in a sentence reading "14 years ago". The International 2026 took place at Gamescom. Counting back, the interview is anchored to roughly 2026. That is arithmetic inference from the article's own wording, not a disclosed fact.

Context: a tool entering a room with no house rules

Jack Williams appears as the representative of iTero, a coaching tool built on artificial intelligence. On the other side stands GIANTX, an EMEA-based esports organisation known to have been formed through the merger of Excel Esports and Giants Gaming, competing inside the LEC system operated by Riot Games. If that background holds, the governing framework for the arrangement sits inside Riot Games' third-party software rules and competitive integrity rules.

That matters, because major publishers take very different positions on how far third parties may touch match data. Dota 2, the only title named in the piece, runs on a cadence of large but infrequent updates. League of Legends updates far more densely. That difference is not a trivial technical detail. It determines the value of every machine-learning model.

When a title holds its meta stable for months, a coaching model built on historical data keeps its accuracy across a long window. When the meta turns over every two weeks, the model's value shifts from solving the meta to detecting the meta shift faster than opponents. The same product, two opposite value propositions. A tool marketed identically across both titles is a signal worth re-examining.

I record that judgement at medium confidence, because the interview names no patch. Here I stop: two independent sources, or one sufficiently strong source, otherwise no conclusion.

Most other dimensions fall outside analysis for lack of data. No bracket, no format, no schedule, no roster. Even the team GIANTX currently fields is not named. That absence is itself a signal: this is a B2B product piece written for organisational buyers, not for viewers tracking game by game.

The history of a boundary that kept moving

Before talking about artificial intelligence, it is worth remembering that esports has already drawn a similar boundary and moved it at least once. Coaches were once barred from communicating with players while a match was live. That ban did not come from any coach doing something wrong; it came from the operator being unable to control the content of that information channel. What was banned was the channel, not the behaviour.

That principle explains almost the entire history of esports governance: when an information channel cannot be supervised, the operator closes it. When the channel reopens in legal form, people build more cameras, more delay, more process. The boundary does not vanish; it is simply marked out with more material.

With artificial intelligence, that principle hits a wall. The channel cannot be closed, because the channel is not inside the match. A model running on a laptop in a hotel room three hours before the game sends no signal to the competition server. It violates no existing clause, because no clause was ever written for it.

The natural position of a tool like iTero is not inside the prohibited zone. It sits in the unclassified zone. And everything unclassified defaults to whoever arrived first.

Analysis: four layers of the same question

The first layer is time. In-game assistance is unambiguously banned across every major title. Pre-match and post-match assistance is barely touched. The window between games in a BO3 or BO5 is the genuine grey zone, because that is when information from the finished game flows into the next game's decisions, and nobody controls the speed of that flow. That is where the real dispute erupts, not in the abstract question of whether AI is permitted.

The second layer is exclusivity. In a closed league like the LEC, every member is a permanent member with no relegation slot. A structural advantage therefore is not competed away across seasons. It compounds. A team holding an exclusive tool for three seasons holds three seasons of data no one else can touch, and that data then feeds the very model that produced it. That loop does not exist in an open system, where weak teams are eliminated and strong teams must keep proving themselves.

Put differently: in an open system, tooling advantage is dispersed by elimination risk. In a closed system, tooling advantage has only one direction: upward. This is the point the interview should be read closely for, yet it appears in neither of the two disclosed section headings.

The third layer is definition. AI-assisted cheating is a phrase without an operational definition. It could mean a model reading the live match data stream directly. It could also mean a model predicting opponent behaviour from public data. Those two behaviours sit very far apart on the competitive ethics scale, yet share one string of words.

In the history of sports law, vague concepts have always caused more damage than wrong ones. A wrong concept can be fixed. A vague concept gets read differently by every referee, and each differing reading becomes a precedent. The trap of 2026 was not in the hand; it was in the belief in a definition that did not exist. When IFAB rewrote the handball law that year, the share of handball incidents handled consistently in the data I collected in Russia reached only about one third. The problem then was not weak referees. The problem was the wording.

Esports now stands before a wording like that. And this time, it spreads far faster than football did.

The fourth layer is copying. Jack Williams raising the prospect of being copied suggests iTero positions itself as the market leader. But in coaching software, the moat is not source code. It is proprietary data and proprietary relationships. Source code can be rewritten in a few months. An exclusive contract with a top team cannot.

This produces a paradox: the more easily a product can be copied technically, the more its value depends on contracts. And when value depends on contracts, pressure shifts from the research room to the legal room. That is the marker of a technology that has entered its commercialisation phase.

We look for justice on the pitch, but what we usually get is merely a good enough excuse to stop arguing. With AI coaching, that excuse has not yet appeared.

What matters is not the technology

Based on my experience following these matches, the biggest controversies in modern sport rarely originate in new technology. They originate in the lag between a technology's arrival and the moment the rules are finished. Technology always arrives first. Rules always arrive later. In between, every act is legal, including acts that will later be called cheating.

Every VAR error is a crack in the mirror that reflects the rulebook. But the crack only shows when light is cast on it. With AI coaching, nobody has switched the light on.

No data in the interview allows me to judge how strong iTero's product actually is. No sample size, no evaluation methodology, no test results. That is a limitation of the source, and I keep it in the piece rather than filling it with speculation. Readers need to know which parts are facts, which are inference, and which are gaps.

One more thing to state plainly. I do not have enough facts to judge the ethics of the iTero and GIANTX arrangement. Its duration, the scope of data, whether there is a clause reopening access to other teams, whether there is an audit mechanism, all of it is absent. Issuing a verdict without the original document is a mistake I once made, and I am not repeating it.

The counterintuitive angle: the frightening thing is not the model

The intuitive reaction to AI coaching is fear of the machine. Fear that it calculates faster than a human, sees the enemy composition in advance, turns the coach into a button-presser. That fear is understandable, easy to picture, and easy to sell.

But in sport, a new device has never been what damages fairness. What damages fairness is the silence of the governing body when that device appears unevenly distributed. A wrong decision does not destroy a match; the silence that follows it destroys trust.

Applied here: if every LEC team could use the same tool at the same price, the AI coaching story would have almost nothing left to discuss. It would become an operating cost, like a gym analysis machine or replay software. Tension arises only from the word exclusive.

And that word was not created by AI. It was created by the structure of a closed league. Artificial intelligence is merely the flashlight pointed into a room that was already dark.

There is a professional paradox here. I once built a player evaluation model and advised against signing a defender, based on his fouls per match. The model was right about the indicator and wrong about the football. The player was signed anyway, and became a pillar of a title-winning side. The lesson I drew was not to abandon models, but to add a limitations-of-data section to every conclusion. A model predicting human behaviour without modelling the teammates around that human is measuring only half the truth.

With AI coaching, that risk is enormous. A model can optimise one metric and ignore the entire context that gives the metric meaning.

One comparison, and I use it only once

Someone working in Korea who follows the AI coaching wave may recognise a familiar pattern from VAR officiating crews. In both settings, a new system is bolted onto an old rulebook, and the old rulebook is expected to handle the new system on its own. Referee committees across Asian competitions went through exactly that arc when VAR arrived: the tool came first, the training of the people using the tool came much later.

VAR was born from the fear of error, but it nurtures the fear of late truth. The more accurate the tool, the harder it is to accept its silence. AI coaching will follow that same road, differing only in speed. VAR took nearly a decade to go from trial to global standard. A machine-learning model can go from a hotel room to an exclusive contract in a single transfer window.

This is also why I look at esports' youth ecosystem with a degree of concern. The transfer market runs on unwritten law, where value is set by expectation rather than achievement. Esports careers are shorter than football careers, while youth development and post-retirement support systems are close to nonexistent. Once AI coaching becomes standard, the pressure on an eighteen-year-old will not be about outplaying an opponent. It will be about outplaying a model fed on his own data. Nobody has prepared anyone for that.

What should be written before it becomes a headline

At the current pace, pressure will come from two directions and arrive sooner than expected. The first is the teams without an exclusive arrangement, once they start losing games whose cause cannot be proven. The second is the publisher itself, once it realises it is permitting a private competitive asset to operate inside a league it owns.

The most reasonable exit is not a ban. Banning a tool that runs outside the competition server is nearly unenforceable, and every unenforceable ban weakens the entire rulebook. The more reasonable exit is mandatory disclosure: which tools are used, which data is loaded, and which windows are closed.

Three things need to be written down: a permitted-tool list, input data limits, and a random inspection mechanism. It does not need to be perfect. It needs to exist.

In every refereeing dispute I have dissected, the root cause was never a corrupt or incompetent referee. The root cause was a clause written for a world that no longer exists. Esports stands before exactly that kind of clause, and it stands before it holding a contract.

If Jack Williams' interview has lasting value, that value is not in the iTero product. It is in the fact that someone inside the industry spoke publicly about the word exclusive before the rules caught up. When the final headline lands on news pages, people will argue about artificial intelligence. But the real crack lies elsewhere, and it was already there.