Meta will replace humans with artificial intelligence to assess privacy and social risks in its products

Meta plans to significantly change the privacy and risk assessment process for its products, replacing most human reviewers with artificial intelligence. Internal documents reviewed by NPR show that Meta aims to outsource up to 90 percent of such assessments to AI systems.

In practice, this means that things like critical updates to Meta's algorithms, new security features, and changes to how content is allowed to be shared on the company's platforms will be largely approved by a system powered by artificial intelligence.

Previously, all updates to Meta’s platforms, including Instagram and WhatsApp, were thoroughly vetted by humans before they were released to the public. However, in the past two months, the company has doubled down on its use of artificial intelligence, which allows it to roll out new features more quickly. Now, development teams fill out a questionnaire about the product, which is then analyzed by AI, instantly identifying potential risks.

Former Meta employees have expressed concerns that automation could lead to increased risks that could lead to negative externalities associated with product changes, as AI may miss serious issues that humans can detect. In response, Meta said that human expertise will still be used to assess complex issues, with AI only handling low-risk solutions.

This change came shortly after the publication of Meta's quarterly content moderation reports, which showed a decrease in the number of removed materials but an increase in bullying and violent content.