Friday, October 24th 2025
EA's AI Tools Reportedly Cost Game Devs Time Even After a Year of Use
Electronic Arts, the gaming giant behind games like The Sims, Apex Legends, and the Battlefield franchise, today announced a new partnership with Stability AI that it claims will help artists more quickly iterate and develop games more efficiently. However, according to a recent report by Business Insider, EA has been pushing its employees across the board to use AI for at least a year. Sources who spoke to Business Insider revealed that, despite what EA says about any AI tools it plans to develop with Stability AI in the future, the current in-house AI tools have caused more hassle than anything else. EA employees familiar with the AI tools say that the AI tools often hallucinate, to the point that developers have to go in and manually fix code that the AI tools generated.
Artists also took issue with the fact that their art was being used to train these AI models, which would theoretically devalue their work and result in reduced demand for artists. There are also reports that around 100 employees were recently terminated from the QA department because the AI could easily review and summarize tester feedback. Despite these complaints, EA is leaning into AI more heavily than ever with the aforementioned Stability AI partnership. News also just broke that Krafton, the publisher behind PUBG and the Subnautica franchise, has announced that it will be pivoting to become an AI-first company. Specifically, Krafton will be spending $70 million on GPU horsepower to power an increased reliance on agentic AI for automation and efficiency. Towards the end of September, it was revealed that EA's new owners—a group of investors—would be leaning on AI in order to "significantly cut operating costs," which has clearly already started to materialize.
Sources:
Business Insider, Krafton, PCGamer
Artists also took issue with the fact that their art was being used to train these AI models, which would theoretically devalue their work and result in reduced demand for artists. There are also reports that around 100 employees were recently terminated from the QA department because the AI could easily review and summarize tester feedback. Despite these complaints, EA is leaning into AI more heavily than ever with the aforementioned Stability AI partnership. News also just broke that Krafton, the publisher behind PUBG and the Subnautica franchise, has announced that it will be pivoting to become an AI-first company. Specifically, Krafton will be spending $70 million on GPU horsepower to power an increased reliance on agentic AI for automation and efficiency. Towards the end of September, it was revealed that EA's new owners—a group of investors—would be leaning on AI in order to "significantly cut operating costs," which has clearly already started to materialize.

20 Comments on EA's AI Tools Reportedly Cost Game Devs Time Even After a Year of Use
And I say successfully as MIT report showed 95% of companies that poured $30-$40 billion into AI are getting zero return on investment.
ASML dropped €1.3 billion into the French Mistral AI, they gotta make it worth it, cause that's some dough right there.
I am not least surprised EA has been trying for a year to make it work, but the ball is in the management team here, not the blue/white collar dude making the games.
It being use in coding I don't have an issue with, it's just another tool. Yes, there are plenty of doom-laden reports because it's very early days, so obviously there are teething problems, and also riding the AI-hating bandwagon is very trendy and gets you social media clout. But most likely in 5-10 years everyone will still virtue signal online but adapt IRL.
The use of AI for any art-related stuff is a tragedy though. Yeah, some will say it's the same as how CGI replaced other stuff, but that was and still is a tragedy too, plus it;s kinda different.
EA is still probably going to charge $70 for their AI generated slop games.
Its either malicious compliance from the devs who know and are trying to slow the rate of the oncoming onslaught or maybe EA/M$ are doing what they normally do and fire all the devs that know too much about their job and thus cost too much so they can hire an intern straight from college that doesnt know shit about the system and gets paid in food stamps... Similar to Ubi soft. Full of interns that have to part time as an uber driver/doordasher to make ends meet because Ubi arent paying them enough and now theyve also all been fired...
www.techpowerup.com/342155/ubisoft-layoffs-to-slash-jobs-at-two-studios-with-cost-cutting-in-mind
The technology is fundamentally flawed...
AI will not replace you, but a person that knows decent prompt engineering might. AI is here to stay. At the very minimum, Copilot assistant is taking the notes and summarizing the tasks while we are on call in Teams. If you haven't used Copilot (or other engine) to prepare a mass email to 10+ managers you should....I mean you really should. There is no beating about the bush, if you want to be up to date with current reality, you have to be using AI in your daily workflow. Otherwise you will be left behind....
Maybe we should hold off on AI for now
The bottom line is this: AI might be able to implement your logic faster, but also comes with an error margin. Now... implementing business logic in code is without an error margin, except you cannot catch any anomalies that escape this logic - for those cases, you want to implement some sort of acceptance rule (these are insurance products) to make sure some person can take a look at it and do manual acceptance.
So what have you got in the end?
- Business logic: 95% gets handled without error by logic and 5% is an anomaly that requires personal attention. You can measure and check exactly what is and what is not caught by business logic, and you fully understand the environment.
- AI: 95% gets handled in ways you can't really make visible but the end result is 'satisfactory' but might contain anomolies, and the other 5% is still not handled correctly enough so you still have to check it as well. Ergo you might need some 5% capacity like you would wrt business logic in human workforce, but you ALSO need another 5% in workforce to maintain and improve and double check the AI. And that is considering the AI can actually reach a 95% success rate - in most applications even that is not what businesses look for, they want 98% or more, or there is no actual gain, no money was saved, the business case just won't fly.
This is a massive bubble and it is going to burst.
After that bursting sure there are applications where AI is an improvement. But the vast majority of companies don't view it like that. They want AI to cut costs everywhere.
And that is not even considering the mid-term implications of people getting wronged by an AI, starting a court case over it, and then we're going to deep dive into 'how did this AI get to this rationale' because it resulted in actual damages for an actual person. And the answer is... we just fucking can't, because we just ran an OpenAI model with our own inputs. So again: this is going to be a major fucking shitshow if companies actually choose to completely rely on AI to do work for them. You're gonna need humans. And we're gonna need a lot of em, because as AI usage proliferates, so too must the checks and balances around it, if companies want to keep their feet dry.
Its all one big paradoxical affair. A venture into pointless improvements that will backfire. Similar things occur in the LLM development itself: it is slowly starting to dawn on researchers that brainrot even occurs within AI models themselves, as they seek shortcuts to earlier thought patterns, because obviously there is no point in rethinking the same thing again. Its so incredibly logical and basic, it baffles me why this hasn't been a thing from the get go. Of course, the model will consolidate on its own knowledge. Input = output. Its the same thing that happens to humans.
I don't buy EA games anyway, so I'm not affected.