Amid CEO Mark Zuckerberg's full-scale bet on artificial intelligence, Meta has recently made significant adjustments in its internal operations. To optimize operational costs and improve processing efficiency, the company is accelerating its "AI review" strategy, aiming to gradually replace human moderators with large language models.

According to the latest disclosure by the Financial Times, Meta has deeply integrated artificial intelligence into its platform's content and advertising compliance review processes. Data shows that currently, about 50% of content review requests on Meta's platform are handled automatically by large language models. The company plans to further reduce the proportion of human intervention by the end of this year; in the review of certain types of violations, the involvement of human moderators could even be reduced by more than 90%.

For a long time, Meta's control over platform content mainly relied on a hybrid model of "automated systems + human review," which also included a large number of third-party contractors in the human team, and user appeals for penalties were usually handled by humans. With the maturation of large model technology, this model that relies heavily on human screening is being broken.

This adjustment is backed by Zuckerberg's substantial investment in AI infrastructure and talent recruitment. He is committed to building what he calls a "personal super intelligence," which not only serves content compliance but also aims to provide users with highly customized AI products and intelligent agent services.

For Meta, this transformation is expected to save the company billions of dollars in operational costs annually. Although this strategy demonstrates efficiency on the technical level, whether AI review can completely replace human judgment in complex contexts and controversial cases remains a focus of external attention. As the end-of-year goal approaches, this "AI takeover" plan for platform governance is entering a deep implementation phase.