EA Says AI Now Handles 85% of Quality Assurance Testing

EA says machine learning now handles roughly 85% of its quality assurance work, a figure that cuts straight to the heart of what Sims players feel every patch day: whether fixes land cleanly or bring new problems with them. Andrew Wilson made the comment at the private iicon event in Las Vegas, putting a hard number on how deeply AI has moved into EA’s testing pipeline.

Wilson described the system as doing the repetitive work that bogs down large-scale testing, including powering a game on and off, checking for crashes, and running other mechanical verification passes. His point was not that human testers are gone. It was that EA wants people focused on the judgments machines cannot make, sorting through results, flagging real problems, and deciding which bugs matter most in a live game.

That distinction matters for The Sims 4, where updates often solve one issue while exposing another. Players know the pattern well: a patch aimed at stability can still leave broken UI elements, odd CAS behavior, or new performance hiccups behind. EA and Maxis have already acknowledged that pressure by rolling out a quality-of-life roadmap in September 2025 focused on stability, performance, and community feedback. In November 2025, EA said a Sims 4 update would include more than 150 community-voted bug fixes, a sign that the studio is still trying to close the gap between what gets tested and what reaches actual households in live play.

Wilson’s remarks also fit into EA’s broader AI push. On October 23, 2025, EA announced a partnership with Stability AI to co-develop generative AI models, tools, and workflows aimed at helping artists, designers, and developers. EA said the work would unlock faster iteration and expand creative possibilities, while Stability AI said the goal was to reimagine game development.

Still, the public message from EA does not entirely match some employee accounts that have surfaced around the company’s internal AI rollout. Reporting has described staff members saying AI tools produced flawed code and hallucinations that required manual correction, even as leadership urged roughly 15,000 employees to use AI for a wide range of tasks. That tension is the real issue for Sims players: if AI is handling most of the repetitive QA load, the quality of the remaining human review will determine whether edge cases get caught before they hit live saves, modded households, and the rest of the game’s messy, always-online reality.