Meta has replaced roughly half of all human moderation requests with large language models in 2025 and plans to push that share above 90 percent for some content types by the end of the year, saving billions annually according to the Financial Times. Meta disputes the cost argument, citing tests since March showing its language models make 13 percent fewer errors than humans when enforcing content policies while catching 10 percent more actual violations. Unlike traditional ML classifiers, the models are supposed to better grasp nuance and cover more languages. However, employees warn the rollout is too fast, with one insider stating the models still remove or shadow-ban harmless content and there isn't enough oversight. The transition is leading to layoffs, especially among external contractors. The Financial Times also reports a model swap: Meta had been using Google's Gemini for moderation and support but recently told staff to switch to its own new foundation model called Muse Spark, which is trained on past decisions made by human reviewers.