Major League Baseball (MLB) has decided that managers should continue striking out on their own, with no algorithm allowed to pinch-hit for questionable judgment. The US professional sports league has effectively banned teams from using generative AI on league-issued dugout iPads during games. The midseason crackdown targets custom applications that some clubs were using to analyze live game information and recommend substitutions, pitch selections and other tactical decisions traditionally left to players and coaching staff. MLB executive vice president of baseball operations Morgan Sword warned teams about the change in a June 11 memo sent to general managers, assistant general managers and video coordinators, with the restrictions taking effect on July 15, just before the second half of the season began.
According to reports, teams had pushed MLB’s technology rules further than intended by installing custom applications capable of feeding ongoing game data into generative AI systems. Those systems could then attempt to predict the next pitch, recommend what a pitcher should throw or dynamically update a team’s game plan as a lineup turned over for the second or third time. In other words, the iPad was no longer simply showing coaches what had happened. It was beginning to tell them what to do next. As many as a third of MLB’s 30 teams may have used dugout iPads for at least one of these expanded purposes, with pitch calling reportedly at the center of the dispute. The same technology could also have influenced bullpen moves, defensive substitutions and hitting approaches.
The tablets themselves are not being banned, however. MLB-issued iPads currently include Statcast information, approved video angles, automated ball-strike system data and, a custom section where teams could install their own software. MLB has now removed that custom application channel from in-game service. The league first introduced iPads into dugouts and bullpens in 2016, giving players and coaches access to scouting reports, analytics and video without requiring someone to sprint down the tunnel carrying a binder. Access was expanded in 2021 to include in-game video, although footage is delayed and the devices remain locked down and monitored by the league.
The league’s nervousness is understandable. Baseball is still living with the aftermath of electronic sign-stealing scandals, most notably the Houston Astros affair, in which technology was used to decode opposing catchers’ signals and relay pitch information to hitters. Allowing a generative AI system to consume live game data, identify patterns and predict the next pitch risks creating the same competitive-integrity problem with considerably more sophisticated software and fewer conveniently audible rubbish bins. In its review, MLB said it found no evidence that any club had violated existing rules covering electronic devices or sign stealing. That makes the decision appear preventative rather than punitive. The memo also reportedly did not establish specific penalties, suggesting league officials concluded it was better to shut down the practice before someone built a system capable of crossing the line faster than MLB could write another rule.
The decision has reportedly irritated some analytics and research staff, particularly those employed to find precisely this sort of competitive advantage. From their perspective, AI-assisted pitch calling is simply the next stage of data-driven baseball. Models can ingest pitch type, location, count, batter tendencies and previous sequences far more quickly than a coach armed with a clipboard and several decades of instinct. That argument is unlikely to reassure the commissioner’s office, because there is a considerable difference between using AI to prepare a scouting report before a game and having an algorithm recommend the next slider while the batter is standing at the plate. Once software begins influencing decisions in real time, the contest risks becoming partly a battle between proprietary models, computing resources and whichever club has hired the cleverest collection of data scientists.
For now, AI can continue helping teams prepare, model player performance and analyze the endless stream of information produced by modern baseball. Once the first pitch is thrown, however, the machines will have to remain on the bench, leaving managers free to call the wrong reliever entirely by themselves.
