'Mad Cow Disease': Grand Theft Auto Dev Warns That Generative AI Could Harm Games

In brief

Rockstar Games co-founder Dan Houser warned that AI models scraping AI-generated data could trigger a creative breakdown in games.
He warned that training AI on synthetic data would produce models that “mirror themselves” and degrade over time as original human work becomes harder to find.
Houser said he “couldn’t see how the information gets better” as studios adopt tools built on increasingly self-referential datasets.

Rockstar Games co-founder Dan Houser—one of the key figures who helped propel the popular Grand Theft Auto franchise—delivered a stark warning about the direction of generative AI in video games, arguing that models trained on synthetic data will eventually degrade quality across the industry.

Speaking on Virgin Radio UK, Houser said AI models that are scraping an internet increasingly populated with AI-generated content risk pushing the medium toward a recursive loop and collapse in quality.

“As far as I understand it, which is a really superficial understanding, the models scour the internet for information, but the internet’s going to get more and more full of information made by the models,” Houser said. “So it’s sort of like when we fed cows with cows, and got mad cow disease.”

His comments come as generative AI increasingly becomes a part of video game development. A recent Google Cloud survey of 615 developers found that nearly nine in ten studios already use AI agents somewhere in their pipelines, with many saying those agents now influence live gameplay through real-time NPC behavior, tutorials, and automated testing.

Experts, including Google CEO Sundar Pichai, argue that as original human-made material becomes harder to find online, development will struggle, and AI models will be forced to train on synthetic data produced by other models. That loop, Houser said, will eventually cause the information to degrade.

“I can’t see how the information gets better if there’s not… they were already running out of data,” Houser said. “Algorithms will become saturated by the definition of how they are sourced and how they are therefore constructed, and it’s going to become this sort of mirror of itself.”

He also took a dig at executives pushing generative AI, suggesting that they “maybe aren’t fully-rounded humans.”

Houser left Rockstar Games in 2020 after more than two decades shaping blockbuster franchises like Grand Theft Auto, Red Dead Redemption, and Max Payne.

While Houser warned about the risk of quality information drying up online, developers are using AI to speed up production, with teams relying on agents for coding, localization, playtesting, and real-time NPC behavior. Small studios said AI helped them compete, while larger publishers struggled to adapt.

“If you’re not on the AI bandwagon right now, you’re already behind,” Kelsey Falter, CEO and co-founder of indie studio Mother Games, recently told Decrypt.

Jack Buser, global games director at Google Cloud, warned that studios that are unable to adapt may not survive the shift.

“Some of these game companies are going to make it, and some of them are not,” Buser told Decrypt. “And some are going to be born through this revolution.”

Major publishers like Ubisoft, Square Enix, Electronic Arts, and Krafton have recently announced major generative AI pushes, including integrating AI-powered tools and revealing experimental game projects. Such moves have come following widespread layoffs across the video game industry over the last couple of years.

Despite his warning, Houser said he still found the technology fascinating.

“I’m slightly obsessed by the fact that when you search for the same thing again, it doesn’t give you the same answer,” he said. “It’s wrong a lot of the time, but it says it so confidently.”