The internal chatbot of Electronic Arts (AE) is called ReefGPTIt's supposed to write code, speed up AI game development, and free up creative time. In practice, however, it generates errors that developers must fix manually, making more effort instead of doing less. This happens just as EA announces a strategic partnership with AI stability To bring artificial intelligence to the heart of video game production. The stated goal is noble: more realistic textures, 3D environments generated by text prompts, faster workflows. The problem is, those working on the ground tell a different story. Developers worry they're training their replacements, while management is ramping up automation to manage a $20 billion debt burden.
The partnership that promises everything
As I told you at the beginning, Electronic Arts and Stability AI have signed an agreement to co-develop generative AI tools for artists, designers, and developers. Stability AI is the company behind stable diffusion, one of the most widely used text-to-image models in the world. The first practical application concerns the Physically Based Rendering materials, ultra-detailed textures that make video game surfaces look realistic under any lighting conditions. It's useful for making a soccer jersey shine under the stadium floodlights or capturing the glow of a coffee table in the morning light.
The ambition, however, goes much further. EA and Stability AI want systems capable of pre-visualizing entire 3D environments starting from simple text descriptions. Type “rainforest at sunset with ancient ruins” and the system should return a complete game environment. Rick Stringfellow, head of visual content at EA, talks about “evolving the craft of game development” and “tools to tell deeper stories.” Prem Akkaraju, CEO of Stability AI, adds that "innovation starts with the creator" and that progress is only possible when scientists and creatives work side by side. Is it all right, is it all beautiful, is it all true?
But then there's ReefGPT
Behind the scenes, the reality is different. EA's internal chatbot, designed to speed up development, writes code with systematic errors. Generates assets that require corrections. Produces what developers call "hallucinations": output that seems plausible but is completely wrong. According to internal reports, the result is the opposite of what was promised. Instead of reducing the workload, AI increases it. Developers spend hours fixing what the machine did wrong.
Friction between management and development teams is growing. On the one hand, executives see artificial intelligence as the solution to reducing costs and accelerating production. On the other, those who write code every day... it feels more and more like a "caregivers" of the system that could soon replace it. As one EA employee told Business Insider, “AI is a productivity tool that often creates more work, not less.” The widespread feeling is that you are training your replacement while correcting every error in the machine.
A study published in May 2025 of 42 game design professionals highlights the paradox. While generative AI accelerates prototyping and ideation, it raises concerns about originality, creative dependency, and the devaluation of human-made content. Designers fear that automation will reduce opportunities for juniors, especially in roles that traditionally involve repetitive asset creation or initial iterations.
The weight of debt
The context makes everything more pressing. EA It was just purchased for $55 billion in what represents the largest deal in gaming history. Yes, because the operation brings with it a debt of 20 billion dollarsThe new owners are focusing on artificial intelligence to cut operating costs and manage this financial burden. The partnership with Stability AI is part of this expense-reduction strategy.
The approach is not unique to EA. Take-Two Interactive e CD Projekt They acknowledged in their investor documents that the use of AI poses reputational and legal risks. Michal NowakowskiCD Projekt's co-CEO, called generative AI "quite tricky when it comes to intellectual property." You can write a nice prompt and end up with a nice game environment, only to find that the AI has stolen it by "taking inspiration" from some successful competitor's title. For this reason, The Witcher 4 It won't use these solutions. But while some are proceeding cautiously, EA is accelerating.
The numbers tell another story
A Google Cloud research A survey of 615 video game developers reveals that 90% already use some form of artificial intelligence in their workflows. 97% believe AI is transforming the industry. But the data also shows that developers use it primarily for support tasks: automated playtesting, chat moderation, and code improvement. As highlighted by recent analyses, content creation remains primarily human territory, especially in larger studios
The gap between executive enthusiasm and developer skepticism is clear. 85% of executives say they use AI, compared to 58% of artists. The difference isn't technical. It's a question of perspective. Those in charge see efficiency and cost reduction. Those working in the field see mistakes, corrections, and anxiety about the future.
The question that remains
EA claims that AI won't replace humans but will "supercharge" creativity. It says that machines can sketch, generate, and analyze, but they can't imagine, empathize, or dream. That remains the job of artists and storytellers. But when a system generates errors that require hours of correction, when it forces developers to act as quality controllers for the machine's output, when the internal climate talks about training its replacement, that distinction sounds optimistic.
AI games promise speed, efficiency, and larger worlds. In the field, they offer controlled chaos, increased workload, and an uncertain future. As Digital Trends notesEA is betting on artificial intelligence to power the future of video game development.
But inside, the revolution already resembles a slow-motion revolt.
