Jack Williams, iTero and the Legal Vacuum of AI Coaching in Esports

The eighth minute of the break between game two and game three. In those eight minutes, what is each team allowed to do?

If a piece of software has just finished reading your opponent's last two hundred matches, has identified that they win 68 percent of games in which they control the first major objective, and has proposed an optimal draft plan — is that legal analysis, or is it an illegal in-game assistant wearing the costume of software? No league on earth currently has a clause written clearly enough to answer that question with a yes or a no. That is precisely the territory Jack Williams, the man behind iTero, is staking out.

I read the match report before I read the news, because a match report does not know how to lie. But a match report can only record what the rules have already defined. When new tools arrive faster than the rules do, the report falls silent — and that silence is the real problem.

Context: one interview, three names, and a question nobody has answered

The Jack Williams interview revolves around three entities: Williams himself, the iTero tool he represents, and the organisation GIANTX. The article's public content divides cleanly into two parts. The first concerns iTero working exclusively with GIANTX, together with the likelihood of competitors copying the model. The second concerns the risk of artificial intelligence being used to cheat.

These are two very different frames for the same object. The first is a commercial frame: how do you protect your competitive advantage when your product can be cloned within months. The second is an integrity frame: how do you ensure that the tool does not become a vehicle for breaking the rules. But sitting between the two is a third frame the interview barely touches — the league-fairness frame. That is the frame I care about most, as someone who works in the rules business.

GIANTX, according to what the EMEA esports industry records, is an organisation competing within the LEC system — the top-tier European league for League of Legends, operated by Riot Games. The organisation is widely understood to be the product of a merger between older esports teams. If that is accurate, the governing framework for the iTero–GIANTX arrangement is not Valve's rulebook but Riot Games' team rules and competitive-integrity regulations. This is an important detail, because the two largest Western esports publishers have historically taken very different approaches to third-party tooling.

I say "if that is accurate" deliberately. In seventeen years of observing this industry, I have learned one thing: nothing is more dangerous than treating a reasonable inference as a verified fact. An organisation's name, a positioning line, a merger once reported in the press — all of these are pieces that may be correct, but they are not yet enough to conclude. In this article I will be explicit about which parts are facts, which are inferences, and which remain open.

The second name to place is iTero. It is a product described as an artificial-intelligence-based coaching tool. No public documentation in the material supplied to me shows the product's evaluation methodology, its sample size, or any performance metric whatsoever. That means every assertive statement about iTero's power — whether from its founder or its partner — belongs to the category of claim, not the category of evidence. For a writer who works from match reports, that is a note that must be entered at the top.

And Jack Williams? He appears as a representative and a product builder, not as a competing player. This shapes the entire reading of the interview: it is an exchange between a tool vendor and the public, not an exchange between a player and the fans. The tone will be a salesperson's tone, even when the speaker is at their most honest.

The meta that is actually changing is not inside the game

When analysing an interview about a coaching tool, the first instinct of anyone in my profession is to go looking for the patch. Which patch is shaping the meta? What does the pick-ban rate look like? How have win rates moved?

In this case I have to say it plainly: there is no patch in the source material. No version number, no balance change, no pick-ban data, no map, no item. Any in-game meta commentary would be fabrication. But there is another meta layer that can still be analysed, and it is the layer the interview is actually talking about: the meta of the tools themselves. That is, the way teams solve a patch is itself shifting.

Here, the difference between titles becomes a first-order commercial variable. Take the two poles.

Valve, the publisher of Dota 2, is famous for infrequent but seismic updates. It can leave a meta stable for months, then release a large patch that shuffles almost everything. In that environment, a machine-learning model trained on historical data retains its value for longer windows. The tool's value lies in the depth of its historical modelling.

Riot Games, the publisher of League of Legends, goes the opposite way with a dense update cadence. Every two weeks, a new patch. Every patch, a few numbers adjusted. In that environment, the half-life of any pattern an AI learns is sharply shortened. The tool's value shifts from "solving the meta" to "detecting the meta delta faster than your opponent".

This is the key point most fans miss: the same AI product, placed into two ecosystems with different patch cadences, will have its value inverted. In one it is a library, in the other it is a radar.

If iTero is marketed identically for both environments, that is a red flag. Not a red flag of dishonesty, but of positioning that has not been validated per title. A tool sold to a Dota 2 team and a tool sold to a League of Legends team must, technically, be two different products. If they are the same, then either one is mispriced, or both are.

And this is where I have to plant a data red flag. There is no information at all in the source about the patch rhythm iTero was designed to serve, about whether tournament servers lock versions, or about the data window the tool is permitted to access. This is the single largest analytical gap in the entire subject. You cannot evaluate an analytics tool without knowing what it analyses, under what conditions, and with what data.

Exclusivity: a commercial decision, a competitive consequence

The first public part of the interview deals with iTero working exclusively with GIANTX, and with the likelihood of the arrangement being copied.

At the commercial layer, the story is simple. A startup signs an exclusivity deal with a team. The team gets a tool no rival has. The company gets an anchor client, a case study to sell to the next client, and an exclusive data stream to improve its product. This is a textbook B2B model.

At the competitive layer, the story is very different. In a closed league — a franchise model like the LEC — all members are permanent members. There is no relegation. No dropping out of the system. That means a structural advantage held by one member will not be competed away over time; it accumulates. In an open system with promotion and relegation, weak teams are culled and the advantage is dispersed. In a closed system, it freezes.

In other words, the same exclusivity arrangement, placed in an open league versus a closed league, produces two entirely different levels of unfairness. This is the comparison I want to stress, and it is also the comparison most commentary on esports tooling skips.

Here I must state my confidence level clearly. I have no evidence that the iTero–GIANTX arrangement includes a clause forbidding the company from selling to other teams. "Exclusive" can mean the tool is used by only one team, or that it is sold only in one region, or that the company only partners deeply with one counterparty for a set period. Those three meanings lead to three different legal consequences. But whichever meaning is correct, the governance question remains: if a tool materially affects competitive outcomes, the league operator will face pressure either to mandate equal access or to restrict the tool.

VAR is not wrong. The people operating VAR are only people. And so are league operators. They will not actively go looking for a fairness problem nobody has complained about. They react only when there is enough noise.

Esports history gives us a very recent precedent. In-match communication between coaches and players was progressively tightened year by year, not by a single ban but by a series of small adjustments. First it was permitted. Then the timing was restricted. Then who could speak was regulated. Then which devices could be used to speak was regulated. Each step was small, and together they amounted to a completely different regime.

AI-based analytics tools are on a similar trajectory. They are not banned yet, because nobody has managed to define them. But they are approaching the zone where league operators will be forced to define them.

Cheating: when does a suggestion become a violation

The second public part of the interview deals with the risk of AI being used to cheat. This is the most legally interesting part, and the most ambiguous.

Start with the easy part. Real-time assistance during play is unambiguously banned in every major title. There is no grey zone there. If a tool reads the match state and issues instructions to players while the game is live, that is a violation, full stop.

The real grey zone sits in the between-game window. In a BO3 or BO5, there are breaks. During those breaks, coaching staff are allowed to talk to players. They are allowed to analyse. They are allowed to adjust tactics. So if an AI tool does exactly what a good analyst does, only faster and more thoroughly — what is that?

There are three possible positions, and each leads to a different legal conclusion.

Position one: AI is a tool, like a spreadsheet. The user must interpret the output. Responsibility lies with the human. Under this position, nothing is illegal.

Position two: AI is a coaching-staff member. It issues recommendations in a form a human can follow immediately. Under this position, you must distinguish between a suggestion that makes a human think and an instruction that makes a human execute.

Position three: AI is an unlicensed coach. If a league caps the number of coaching staff permitted on site, then an automated system playing an equivalent role may be circumventing the headcount rule.

What is striking is that none of these three positions is written explicitly in any esports rulebook I have ever read. Current rulebooks were written for a world in which analytics tools are passive. AI forces them to become active.

And this is the central paradox of the whole subject: the more useful the tool, the more likely it is to be treated as cheating. A useless tool raises no ethical problem at all.

There is a test I use in my work, and I suggest league operators adopt it. I call it the match-report test. Imagine a referee standing beside a team, recording everything that happens during the break. If that behaviour, fully recorded, would still leave you comfortable publishing it to the public — then it is legal. If you feel the need to hide it, there is a problem.

This test is imperfect. It cannot be quantified. But it is a better starting point than silence.

I also have to concede something I often hear managers say, and this time I partly agree with them: the pace of tool development outruns the pace of rule-writing. A regulation written today may be obsolete in eighteen months. But that difficulty does not excuse the obligation to write. It only means you must write in a way that can be updated.

A counter-intuitive angle: fans do not need rules, they need a story

Before I push back, I have to concede one thing. Fans' emotions are valid data. When a viewer watches a BO5 and sees their favourite team lose because the opponent has a tool their team does not, their sense of injustice is not irrational. It reflects something real: match outcomes are coming to depend on factors outside the field of play.

But that emotion, however valid, leads to a wrong conclusion if used as a basis for rules. Fans want the tool banned. But banning the tool is neither feasible nor necessary. Not feasible, because data-analytics software is a universal category of software and cannot be controlled. Not necessary, because the problem lies not with the tool but with access.

This is the counter-intuitive part. If you ban AI analysis, you do not eliminate the advantage. You merely transfer the advantage from the team that has the tool to the team that has more good analysts. And the team with more money always has more good analysts. You do not level the playing field; you change its shape.

Conversely, if you force every team to have equal access to the same tool — through a collective licence, say — then you genuinely flatten the field. But then you face a different problem: the league publisher becomes a tool vendor. And once the operator both runs the league and sells the tool, its neutrality on the field is called into question.

I learned this from a small event in my own career. In 2026, when I was a data-editing assistant in Marseille, I reviewed the tape of a match and found that the referee had missed a situation the rules described very clearly: a defending player deliberately touched the ball before it reached the striker's feet. The goal was disallowed for offside. Under the rules, that decision was wrong. I wrote an analysis with situational diagrams, and it drew several times the site's usual readership.

The lesson I drew was not "the referee was wrong". The lesson was: people do not need rules to feel injustice; they need rules to fix it. And rules, to fix anything, must be specific enough that a referee can look them up in three seconds under pressure.

The offside line has never been straight; it is only that today I noticed it bend. The same is true of AI. The boundary between "tool" and "coach" has never been straight. But only when someone looks closely enough does its curvature become visible.

Proposed clauses: fix what can be fixed

A disallowed penalty can be fixed; a legal vacuum cannot. But a legal vacuum can be narrowed, one clause at a time. Here are three proposals I consider feasible at low cost and low risk.

Proposal 1 — Tool disclosure clause. Every team competing in a league must declare to the organisers a list of every data-analytics tool used in preparation and competition, with the supplier and the purpose of use. The declaration need not be published to the public, but must be filed with the integrity office. Cost: near zero. Benefit: organisers get a real map of the tools in circulation, instead of guessing.

Proposal 2 — Temporal boundary clause. Analytics tools may operate pre-match and post-match. During the match, they may supply raw statistics only, without action recommendations. Specifically: win rates may be displayed, but "you should ban champion X" may not. This boundary is enforceable because it sits in the software's output format.

Clause 3 — Equal access clause. If a top-tier league permits one member team to sign an exclusive deal with an analytics-tool supplier, that supplier must offer an equivalent licence package to the other member teams, at a price the league confirms as fair. Other teams are not obliged to buy. Only the right to buy must exist.

These three proposals do not fully resolve the problem. They do not define the entire grey zone. But they achieve one important thing: they convert silence into a database.

A 38-point checklist did not save the season, but it saved the referee's good name. That is what I learned when building a protocol for matches without spectators during the pandemic. A checklist does not make a referee better. It only makes a referee able to explain what they did. And in an industry where trust is the asset, the ability to explain is everything.

Seen from two markets: one line, two readings

There is a reason I look at this problem through slightly different eyes than most Western colleagues. I was born in China and practise in France. I have watched two esports ecosystems react to the same kind of refereeing controversy, and the difference between them is striking.

In Western markets, the first reflex in a new controversy is to go find the existing regulation. If no regulation exists, they argue over principles until someone writes a clause. The process is slow, noisy, but systematic.

In Eastern markets, the first reflex is usually to go find community precedent. If there is no precedent, they wait for a statement from a credible operator, and the whole industry follows. The process is faster, quieter, but heavily dependent on a few individuals.

With AI coaching, these two reflexes produce two different outcomes. The West will take eighteen months to write a tight clause, while teams keep using tools in the grey zone. The East will have a verbal statement within three months, and teams will self-adjust to it. The second is faster, but produces a different kind of unfairness: the team that reads the operator's intent best gains the advantage.

Neither system is better than the other. There are only two different curvatures on the same line. And it is the comparison itself that exposes both.

A conditional conclusion: what comes next

Based on what has been verified — an interview with Jack Williams, the existence of iTero as an AI coaching tool, the relationship with GIANTX, and two public themes of exclusivity and cheating — I draw the following provisional conclusions.

First, the AI coaching debate will not end with the question "should it be banned". It will end with the question "what is permitted, when, and to whom". This is a three-part question, and any league that answers only one part will keep meeting controversy. Confidence: high.

Second, an exclusivity arrangement between a tool supplier and a team in a closed league will become the next major governance topic. Not because it is especially wrong, but because a closed structure turns a temporary advantage into a permanent one. Confidence: medium.

Third, current esports rulebooks will have to rewrite their definition of coaching-staff membership. When an automated system can play a role equivalent to an analyst, any cap on headcount becomes meaningless unless it accounts for tools. Confidence: medium.

What remains open, and it is large: none of us has data on the real effectiveness of these tools. No sample size, no evaluation method, no controlled comparison. Until that data exists, every claim about AI's power in esports — whether a sales claim or an objection — is flying through the air.

The match does not end with the whistle; it ends when people finish reading the report. And the report on this story is still being written. Our job, as observers, is not to predict the outcome. Our job is to ensure that when the report closes, it records what actually happened — no more, no less.