MLB banned AI from the dugout — because the line between preparing and cheating had started to blur

Sports 3 min 80 sources

Baseball quietly outlawed teams from using artificial intelligence to help make live in-game calls. No one was punished. The league acted because a tool had made an old, unwritten line stop meaning what it used to.

Key takeaways

MLB banned teams from using AI on dugout iPads to help make live in-game calls, switching the software off mid-season even though no rule had been broken.
The old line — prepare all you want beforehand, but the in-game decisions are the humans' — only held because real-time analysis used to be slow; AI erased that delay.
The ban protects more than tradition: it stops a spending race where the richest team's model would quietly out-decide everyone else.

Major League Baseball has quietly banned teams from using artificial intelligence on their dugout iPads to help make in-game decisions. A commissioner’s-office memo, issued June 11 and obtained by The Athletic, ordered clubs to stop installing custom apps that fed live game data into AI programs — software that had begun recommending pitch calls, substitutions, and other choices traditionally left to players and coaches [26]. The league switched off the tablets’ “custom tab” on Wednesday night, as the season’s second half began [26][56].

As much as a third of the league had used the iPads for at least one of these purposes, people with knowledge of the system told The Athletic [26]. And no one was punished. The league’s own review found “all clubs were compliant” with existing rules on sign-stealing — teams secretly decoding an opponent’s signals — and electronic devices [26]. MLB acted anyway, mid-season, to stop the practice before it grew. One executive put the logic plainly to The Athletic: “Gotta stop the cheating before there’s cheating now.” [26]

A line that only held because help was slow

The tablets have been in dugouts since 2016 [26]. After the sign-stealing scandals of 2020, the league locked them down — issued by MLB, monitored, no internet, video available only on a delay once an inning ended [26]. Over time teams pushed for more, and the league let them add a “custom tab” for their own data: matchup notes, defensive positioning, the digital version of an old paper binder [26].

The whole arrangement rests on one unwritten line. You can prepare with all the data you want before the game — but during the game, the decisions are the humans’. That line held for a simple reason: doing real analysis in real time used to be slow. Generative AI erased the delay. Some teams began live-scoring the game and feeding it to a model that updated the plan pitch by pitch [26]. The Miami Marlins pioneered calling pitches from the dugout rather than leaving it to the catcher; as many as six teams now make what some call “pitch suggestions” from the bench [26]. Once the machine can both prepare and decide in the same instant, “before” and “during” stop being different things.

Who it lands on

The club reportedly at the center was the New York Mets. Former Met Adam Ottavino said on his stream that the team ran an expensive AI program to help pick pitches, and that owner Steve Cohen had spent heavily on it [45]. The Mets neither confirm nor deny it: interim manager Andy Green said only that the club is “fully compliant” and that MLB makes the rules [45][49]. The league and the players’ union declined to comment [26].

There is a money story underneath the integrity one. AI-for-strategy is a spending race in disguise. A team like Cohen’s Mets can pour cash into software the way it pours cash into players; a smaller-market club cannot. Banning the tool mid-season does more than protect “the human element” — it keeps a sharp, cheap manager competitive against a rich team’s model [26].

What’s still moving

For now, teams are back to running their numbers before first pitch. Static, pre-game information is still allowed — but MLB warned that anything uploaded stays subject to review [26]. The tool did not disappear. It stepped back one pace, into the preparation, and it will push at the line again. As The Athletic put it, fans will still get to blame the human managers for a bad call — “at least for now.” [26]

02 · Lesson · why it matters

The rule that says "no help" only works while help is slow

Every contest draws a line around what counts as "you" — and that line tracks not fairness, but whatever help used to be too slow to use in the moment.

A ban with no crime

A league office switches off a tab on a tablet, mid-season, and a third of the teams have to change how they work. Nobody broke a rule. The commissioner’s own review said so. And yet the tool had to go — fast, before it spread. “Gotta stop the cheating before there’s cheating now,” one executive said. Read that sentence again. It admits there is no cheating yet. It is a rule written against a thing that is only just becoming possible.

That is the strange part worth slowing down on. Baseball did not catch anyone. It felt a line start to dissolve and moved to redraw it before it was gone.

The line was never about fairness

The whole thing rests on one boundary that sounds obviously fair: prepare all you like before the game, but during the game, the decisions are yours. Study the tape, fill the binder, run the numbers — that is preparation. Then the whistle blows and it is on the humans in the dugout.

But that line was never a moral principle. It was a fact about the clock. Real analysis in the moment used to be slow. You could not, between two pitches, re-run every matchup and get a fresh answer. So “you may prepare but not consult” cost nothing to obey — the consulting simply could not keep up with the game. The rule felt like fairness. It was really just a description of what was feasible.

When the friction vanishes, someone has to choose

Generative AI erased the delay. Feed a model the game as it happens and it updates the plan pitch by pitch. Now “prepare beforehand” and “decide right now” are the same act, done by the same machine. The old boundary doesn’t break loudly. It just stops meaning what it meant.

And when a line stops meaning anything on its own, someone has to redraw it on purpose. That is what the memo was. Faced with a tool that could do the in-the-moment thinking, the league had two moves. Absorb it — and quietly turn the game into a contest over who owns the best model. Or ban it — and keep the game a test of the human read. MLB chose ban, for now. Either way, a person had to decide what the game is for. The tool took that decision out of the shadows and put it on someone’s desk.

The same fight, everywhere the machine can think

This is not a baseball story. Every place that ever said “you may prepare, but not consult in the moment” is having the same fight. The exam that allowed a calculator but not the internet. The courtroom, the hiring room, the newsroom, the classroom essay. Each drew a line that held only because real-time help was slow or obvious. Each is now watching that line dissolve, mostly with no commissioner to redraw it — just a lot of people quietly deciding for themselves.

The shape is identical: a boundary you thought was about fairness turns out to have been about friction, and the friction is gone.

Whose advantage the old line was quietly protecting

Notice who the ban helps. AI-for-strategy is a spending race wearing a lab coat. The team that can pour money into software the way it pours money into players — a Steve Cohen, say — pulls ahead of the club that cannot. Ban the tool and a smart, cheap manager can still beat a rich team’s model. Absorb the tool and the deepest pockets win by default.

So drawing the line is not neutral, and it never was. Every place you set it rewards someone. The pre-game boundary looked like plain fairness, but it protected the human reader against the well-funded machine — and it can protect the smaller club and the sport’s texture at the same time. An arrangement can serve its keeper and shelter those beneath it at once. The trick is to see that it is a choice at all, and to ask who each version of the line is for.

We are all managers now

The reader is inside this, not watching it. The same tool that unsettled a dugout sits in your pocket, and you are already deciding — in your work, your writing, your own judgment — where your “no help” line falls. Prepare with it, or decide with it? You are drawing the boundary baseball just drew, only nobody handed you a memo.

And no single seat sees the whole of it. The coach sees a lost edge. The commissioner sees the sport’s texture. The fan sees a manager to blame. The player sees a job half-handed to a machine. Each is looking at a different sliver of the same shift. The line MLB drew will move again, because the tool did not leave — it stepped back into the preparation and is pushing already. What stays a human call is not settled by the machine. It is decided, over and over, by people choosing what they still want to do themselves — and none of them can see far enough to be sure they got it right.

03 · Lab · your turn

Draw the Line

Play the commissioner deciding how late a team may use AI in a game, and feel that every setting trades sharper calls, a human contest, and who money favours.

04 · Hope · carry this

We have handed hard jobs to machines many times before and still kept the parts worth doing ourselves. The line moves, but people keep choosing, on purpose, what stays a human call.