Using generative AI and LLMs in game development is among the most debated—and polarizing—topics in the industry right now. Increasingly, both players and critics are keeping a close eye on how these tools are being used in games, checking the Steam store pages of every newly announced release to see whether the technology played a role in its creation.
Take-Two, the parent company behind Grand Theft Auto 6 studio Rockstar, had previously maintained a team dedicated to researching and evaluating how AI—across different forms—might support game development. That changed in April, when the company reportedly let go of the entire AI unit.
The move came as a surprise, particularly since many organizations are either quietly or openly incorporating generative AI into their production workflows. That said, most of the employees who were dismissed were not working specifically on generative AI.
The group first began at mobile developer Zynga and was later acquired by Take-Two in 2022. After the acquisition, its responsibilities expanded to cover all of Take-Two. In an interview with GamesIndustry.biz, Dr. Luke Dicken—who previously led the team—revealed that the skunkworks effort launched in 2019, years before the 2022 public release of ChatGPT and the wave of attention that followed for generative AI.
“Generative AI is not something I have ever felt especially enthusiastic about,” Dicken said. “I think it comes with a moral responsibility to be handled as responsibly as we can, while also recognizing that for any major corporation in 2025/2026, refusing generative AI would be the wrong approach—and would likely push away a lot of people.”
Dicken made clear that this remains a complicated subject. Still, he added that “some of the most extreme examples of generative AI are so unsettling that you have to make sure you can respond and push back.”
Instead, he said he would like people to take a broader perspective on AI, not just generative AI. “Five years ago, if you told us you had an algorithm that could significantly improve level generation content in a mobile game,” he recalled.
“At the time, people acted like we were bizarre or strange. Now, though, the attention around AI has created an environment where I can say that AI is what’s going to bring your game into quantum computing, and people would agree: ‘Yes, we want AI in the game,'” he continued.
He believes the upside is that “it has helped people become more receptive to conversations about what older, conventional approaches could have done for them before. They’re now more open to the idea that these tools can exist in practice.”
Unfortunately, with the current surge of hype around generative AI, some people may end up walking away from AI research altogether if the hype fades.
“My concern is that generative AI is poisoning the well,” he added. “I don’t believe there’s enough sophistication and careful thought to sustain the older, traditional approaches. With LLMs, we’ve already fallen into a valley of disillusionment.”