Zuckerberg’s Meta will soon use AI to do what thousands of human moderators do

Meta is reportedly accelerating plans to replace thousands of human content moderators with generative artificial intelligence (AI). The transition comes as part of a broader, company-wide cost-cutting drive as the Mark Zuckerberg-led company is spending billions of dollars on advanced AI infrastructure. Citing sources familiar with the matter, The Financial Times reported that the tech group has fast-tracked the deployment of large language models (LLMs) to review posts and advertisements across Facebook and Instagram – that could save the company billions of dollars annually.Massive shift to automated reviewMeta has relied on a hybrid system of automated filters and thousands of human reviewers to police its platforms for harmful content, and handle user appeals. However, this balance has shifted as the report said that Meta has already replaced roughly half of all human review requests with LLMs this year. The company reportedly aims to push that number even higher by the end of the year, cutting human intervention by up to 90% for specific categories of content.While insiders point to massive financial savings, Meta publicly maintains that the overhaul is about accuracy rather than cutting costs. The company claims that initial testing since March shows LLMs actually outperform people, making 13% fewer mistakes than human reviewers while successfully catching 10% more policy violations.“The point of this work is to improve our enforcement efforts, and we’re deploying these more advanced AI systems once we’re sure they’re consistently performing better than our current methods of content enforcement,” Meta said in a statement.Internal friction and the ‘shadow-ban’ problemThe rollout has sparked concern among some Meta employees, who warn that the technology is being deployed without adequate oversight, the report said. Unlike traditional AI, which relies on rigid keywords, generative LLMs are supposed to understand context. Yet, insiders are reported to say the models still routinely struggle with nuance, sarcasm and evolving internet slang. The report also said that one Meta insider revealed that the AI keeps making frustrating errors, mistakenly deleting harmless posts or “shadow-banning” users, a practice where an account's content is secretly blocked from appearing in public feeds.Furthermore, the sudden shift is already triggering heavy layoffs among external contracting firms, with Meta expected to cancel multiple upcoming contracts with third-party moderation groups, the report said.Get the latest technology news and updates. Download the TOI App.