Xbox and AI: discomfort comes to light
A Microsoft Gaming executive stated that Xbox already uses AI “like all companies in 2026.” His remarks, shared then removed from X, reignite tension between internal experimentation and studio distrust.
At Xbox, AI is no longer just a lab topic: it now fuels public debates, internal stances, and visible tensions in video gaming. On October 19, Brandon Stirrup, Xbox's strategy lead, stated on X that Microsoft uses AI “like all companies in 2026.” His account disappeared the same day, but his statements remain in circulation. The message matters because it shows how deeply AI is embedded in the publisher’s practices, while revealing a fracture line with part of the sector.
Une prise de parole qui a tourné court
Stirrup also explained he created a small football game in two days, “bad but improvable,” with only three hours of actual work. His method relies on “vibe coding,” an approach where the idea is described in plain language while AI produces the code. In the same thread, he said studios reluctant to this tool “have no place in the business anymore.” The phrase was obviously poorly received by the community, which never helps calm an already inflammable thread.
To critics, he replied he was “happy to take the hit if hundreds of people at Xbox can stop pretending.” He also wrote: “if I get fired tomorrow, it will be because I told you the truth no one wants you to know.” He simultaneously claimed to “love Xbox and Microsoft” and to have “spent over ten years proving it.” His LinkedIn page still lists him as employed [to be verified], and neither Microsoft nor its gaming division commented on the thread’s disappearance.
Xbox promet de ne pas noyer ses équipes sous l’IA
This statement comes as Microsoft Gaming tries to balance both sides. On February 20, Asha Sharma, appointed head of Microsoft Gaming replacing Phil Spencer, promised teams not to flood Xbox with “soulless AI mush.” The remark responded to a year of grumbling, between the generative Muse model, poorly received by players, and studios already strained by layoffs. Phil Spencer, before leaving, said he imposed no top-down directives and confined AI to discreet uses, prioritizing moderation.
On October 13, five days before Brandon Stirrup’s thread, Microsoft officially launched Xbox Labs, an AI experimentation unit led by CTO Scott Van Vliet. The company thus proceeds cautiously in discourse but accelerates on tools. It is a delicate balance, especially when the words “AI mush” and “strategy” appear in the same context.
Le secteur avance, mais il se méfie
The figures confirm this ambivalence. The annual Game Developers Conference survey, published late January among 2,300 professionals, shows that 36% of respondents use generative AI, including 30% within their studio. At the same time, 52% consider the technology harmful to the sector, versus 30% a year earlier and 18% in 2024; only 7% see a positive effect. A March survey drops declared usage to 29%, indicating enthusiasm is not linear.
Epic Games, for its part, asked its teams to prepare for a “maximum acceleration by AI,” including for its in-house launcher, partly “vibe coded” as well. The movement is broad but lacks consensus. Between productivity gains and fear of quality degradation, video gaming tests AI without a truly stabilized user manual. And for now, everyone seems to write theirs live.
Des usages qui s’installent, des lignes rouges qui bougent
This issue matters beyond Xbox because it summarizes a broader shift in software and creative production. Generative AI is already used to prototype, write code, speed up some support tasks, and experiment with new internal tools. But its adoption still faces very concrete questions: result quality, model dependency, place of creative professions, and team acceptance. In video gaming, where public perception weighs heavily, every announcement can become a large-scale test.
In this context, Microsoft must manage a dual reality: push its teams to experiment without seeming to force their hand. Brandon Stirrup’s remarks mainly show the topic is no longer only technical. It now touches corporate culture, public communication, and internal trust. And no generative model solves that with a click.