Meta AI Shifts Risk Assessment to AI, Amid Concerns

Updated⁣ June‍ 01, 2025
⁢

Meta is expanding its use of artificial intelligence ⁣to assess potential risks associated with product⁢ updates on platforms like Facebook and Instagram.This shift could see⁤ AI handling up to 90% of risk assessments, according to internal documents.

The move raises concerns about the effectiveness of content moderation, notably regarding violent content, misinformation, and risks to young users. Some current and former Meta employees have⁢ cautioned that AI systems may miss critical issues that human reviewers would identify.

Previously,⁢ human teams reviewed updates and new features before ⁣their release. Now, product teams complete a questionnaire that is than reviewed by the AI system. The system ⁤provides an “instant decision” and outlines requirements to address identified risks before ‍a product’s release.

A former Meta executive suggested that⁢ reducing human scrutiny increases risks, making it⁣ less likely that negative consequences of product changes will be⁣ prevented. Meta stated that “human ⁣expertise” would still be used for “novel and complex issues,” leaving “low-risk decisions” to AI.

This change comes after Meta’s latest quarterly integrity reports, ‍which showed a decrease in content removal following ⁣policy changes earlier in the ‍year. However, there ⁣was a ⁣slight increase in bullying, harassment, and violent ⁣content.

What’s ⁤next

It remains to be seen ⁢how⁣ this increased reliance on AI for ⁤risk assessment will affect content moderation and user safety⁤ on Meta’s platforms. The company will likely monitor the results and adjust its approach as‍ needed.