Meta to Shift From Human Moderators to AI Systems

Meta announced plans to dramatically reduce the number of human content moderators over the coming years and lean more heavily on AI-based moderation systems. The company frames the change as a "transformation" meant to catch more issues faster and support many more languages than current human-led workflows allow.

Details on how many contractors or reviewers could be cut were not provided. Meta currently employs thousands of contract moderators worldwide who review content flagged by algorithms and user reports. The company said people will continue to be involved in designing, training and evaluating the AI, and will handle the most complex, high‑impact decisions such as appeals and law enforcement reports.

Meta has been testing large language model (LLM) approaches for moderation and described early results as "promising." One clear benefit cited is coverage: the new AI tools could handle languages used by about 98% of people online, compared with around 80 languages supported today.

Although Meta said its underlying rules are not changing, the shift could alter how users experience enforcement. Many users already complain that automated moderation makes mistakes and that appeals don’t reliably reach human reviewers. Meta counters that its upgraded systems produce "fewer over-enforcement mistakes" while identifying more severe violations.

As part of the rollout, Meta is introducing an AI-powered support assistant in Facebook and Instagram to help with account issues such as reporting content, managing appeals, resetting passwords and assisting with lockouts in select US and Canada cases. The company expects humans to remain involved in critical oversight roles even as automation expands.

For users and moderators alike, the transition raises questions about accuracy, transparency and workforce impacts as platforms increasingly entrust content decisions to AI.