Grok kept getting basic things wrong about Baldur’s Gate 3, and that pushed Musk to make a fix-first decision that affected millions of users for several days. Elon Musk chose to postpone an xAI update until the chatbot started producing the answers he wanted.
Generative AI is now a permanent part of everyday life. It is no longer limited to professional use, because it has spread into daily routines and many people interact with it regularly, almost as if it were a companion. Grok is one of those systems, and it has accumulated hundreds of thousands of followers on X, the social platform where Elon Musk rolled it out seven months ago. Musk, however, reportedly became deeply frustrated with the tool because it was failing to answer questions about Baldur’s Gate 3 correctly, which led him to delay one of its updates so the system could be reworked internally.
Musk’s video game fixation also shaped how Grok was developed
The details come from a lengthy Business Insider report focused on the making of Grok, while also covering the broader planning inside xAI, Musk’s startup. The Palo Alto company was founded to challenge the biggest players in generative AI, but growing pressure from competitors such as OpenAI reportedly forced repeated reassessments of major decisions. Beyond that direct rivalry, though, the development of Grok was also influenced by an unexpected factor: an intense focus on video games.
At one point, the report describes how Musk, after the first version of Grok had launched, was so dissatisfied with its ability to answer complex questions about the Dungeons & Dragons-based RPG accurately that he stopped a key update and redirected attention toward improving the system. The delay only lasted a few days, but the chatbot’s inaccuracies appear to have annoyed him enough that senior engineers were reassigned from other projects to improve its responses before the update was approved.
What stands out most is that this was not treated as a one-off issue. The same report says Musk displayed a similar fixation with League of Legends. It is unclear whether the intensity matched what happened with Larian Studios‘ RPG, but the AI’s performance on questions related to Riot Games‘ MOBA was also seen as weak. In response, internal teams were formed specifically to improve how Grok handled player prompts, requests, and gameplay-related scenarios.
That obsession, however, appears to have caused internal strain. According to the English-language outlet, by the end of 2025 at least five rooms were operating at the same time, each staffed by teams assigned to different parts of Grok‘s game-related development work. The constant shuffling of resources reportedly generated significant tension inside the company and reinforced a view that Musk’s micromanagement style had become chaotic.
The broader context did not help calm things down either. All of this reportedly unfolded after Musk had been publicly accused of cheating in games such as Diablo 4 and Path of Exile 2, and while people associated with Baldur’s Gate 3 were openly criticizing his promises about AI-generated video games.
Source: 3djuegos
