Video games have introduced us to some of the greatest AI villains ever conceived. In System Shock there was Shodan — imagine Kubrick’s HAL 9000 with a god complex — a space station computer which loses its moral directives and begins conducting horrific experiments on humans. Portal introduced the sarcastic AI GLaDOS, who gloats and taunts the player while leading them through a series of deadly tests. Then there’s the rogue AI Hades in Horizon Zero Dawn, who trains an army of robot dinosaurs to wipe out all life on earth — it’s hard to get more extravagantly nefarious than that.
Yet among the people who actually make games, the attitude towards AI isn’t so clear cut. The latest AI technology is threatening to upend game design just as it has provoked existential reckonings in the worlds of music and film, inciting what one expert described as “a civil war within game development”.
The gaming world has not yet decided whether, in our timeline, AI will be the hero or the villain.
Gaming is deeply intertwined with cutting-edge technology, and developers are always looking for new tools to make their games more sophisticated. The concept of AI is not new in gaming — in fact, the term has been used for decades to describe a set of programmed rules that govern gameplay elements such as the movements and behaviour of non-player characters (NPCs). This has been renamed “classical AI”, to avoid confusion with newer technologies such as machine learning (where programs learn from data to solve tasks) and its buzzy new sub-category, generative AI (which can create text, images and other assets).
Some in the gaming world have been quick to trumpet these technologies’ revolutionary capabilities. Electronic Arts CEO Andrew Wilson says that AI is “the very core of our business” while Jack Buser, director for games at Google Cloud, wrote that the technology will be “the biggest change to the industry of games since the introduction of real-time 3D graphics”. But, given that the industry is quick to embrace a new craze — not long ago it was abuzz with NFT games — is this just another case of hype gone wild?
Proponents argue that AI has the potential to revolutionise games from both the developer and player sides. For makers, they could create visual assets ranging from characters to 3D models; write code, storylines and dialogue; generate whole environments and even synthesise vocal performances without the need for actors. For gamers, the prospect is both potentially more thrilling but more of a distant prospect: a “living game” could use generative AI to adjust landscapes and settings according to players’ tastes and behaviours. An inhospitable desert might transform into a lush forest when you’re in need of calm, you could converse with computer characters using your own natural speech, and a game might adjust its difficulty when you’re struggling in combat encounters.
A series of recent prototypes has excited AI evangelists. Last year, Ubisoft debuted an experiment known as Neo NPCs, which allows you to say whatever you want to an in-game character who responds in impressively coherent, if generic, conversation. In February, Microsoft introduced World and Human Action Model (WHAM), which has learnt how to simulate a game after being fed thousands of hours of gameplay footage — though to call its recent Quake 2 demo “playable” would be generous. Meanwhile, Google’s Genie can reportedly create a 3D playable world from a single image prompt. None of these tools are ready yet for mainstream application. Tommy Thompson, a consultant and engineer specialising in AI for games, says of Genie’s latest demo: “Wearing my AI researcher hat, this is really impressive work. But wearing my game developer hat, it’s a weird, expensive toy that has no meaningful applications.”
So beyond the research prototypes, how is generative AI and machine learning actually being used by real developers today? One of the most prominent releases is life simulation game inZOI, a competitor to The Sims which uses AI to power some NPC behaviour and includes generative AI tools that allow players to create custom in-game items and clothing. The game has been a smash hit, selling 1mn copies within a few days earlier this month, but player feedback on its generative tools has been mixed. Generative AI shows more promising potential in the indie space. In upcoming title 1001 Nights you play Shahrzad, creating stories together with an AI-controlled King, while Dead Meat is a goofy murder mystery in which you can type or say whatever you like, and interrogation suspects will respond in character. It remains to be seen whether these features will improve games meaningfully or just feel like gimmicks.
Among the big studios, the approach has been more conservative. Microsoft uses machine learning to detect cheating in online games and Call of Duty makers Activision have said they use generative AI to create some in-game assets, but few companies have wholeheartedly embraced the technology into their workflow. This is partly due to the sheer complexity of game development. “Injecting an unreliable or unknown AI into the process is quite intimidating,” says Thompson, adding that the AI sector is moving so quickly and with such volatility that it might not yet seem ready for a video game, which can take more than five years to develop. He suggests they might wonder: “How can I rely on you to be a technology provider if there’s a chance you don’t exist at the time we launch?”
However, he adds that internally many companies are experimenting with these technologies. Julian Togelius, a professor specialising in AI and games at New York University, says he knows of several big developers that are using generative AI for asset creation, game testing and voice acting synthesis, but not disclosing it publicly. This is likely due to fear of censure: while some in the industry are bullish on AI, others remain critical. The CEO of Take-Two, publisher of Grand Theft Auto, has called the entire concept of artificial intelligence “an oxymoron” while Nintendo has vowed not to use generative AI in the games they make. A recent survey of workers in the industry showed that the number who thought generative AI would have a positive impact on the industry had dropped sharply from 21 per cent in 2024 to 13 per cent in 2025.
What can explain this negativity? In some cases, it’s simply because people don’t think the technology is very good yet. “There’s a large gap between what’s being promised and what’s actually possible with the current technology,” says Ken Noland, co-founder of game co-development studio AI Guys. “Generative AI still struggles when it comes to making meaningful player experiences.”
But there are also more long-term concerns. Some are practical: that digital game stores could become flooded with low-grade AI content. That players overusing AI tools could overwhelm servers and push up costs. That an AI-created character can’t currently be copyrighted, which could pose an obstacle to development and marketing. Other worries are ethical: is the training data used for large language models sourced legally? Could the uptake of AI tools lead to further job losses at a time of painful industry contraction?
And there’s the human question. Many developers want to spend their time making things, not reviewing content generated by an AI. At their current level, these tools still need people to guide them and decide what output is good quality. And, as Thompson puts it: “Players still want a human-crafted experience.” Such sentiments have led one developer to create a “No Gen AI” seal to reassure buyers of indie games.
But change is coming. “I think 2026 is when you start seeing AI becoming more pervasive,” Thompson says. He expects to see some developers insert as much AI as possible into their workflow to save costs and time, and cutting jobs as a result. “The market will dictate the future of that,” he notes. It remains to be seen how gamers themselves will react. “I’m interested to see, as more AI-focused games come out: will people care?”
Find out about our latest stories first — follow FT Weekend on Instagram and X, and sign up to receive the FT Weekend newsletter every Saturday morning