Meta says Meta AI outperforms humans in content moderation and security tasks

Meta has announced a global rollout of its Meta AI support, a system designed to handle customer service tasks and enhance content moderation across its platforms. Meta says the AI tool has demonstrated superior performance compared to human moderators in various experiments, as reported by The Register.

The Meta AI support system is designed to manage tasks such as password resets, explaining content takedowns, and processing appeals. More advanced AI systems are being deployed to identify and remove severe content violations, including scams and illegal content.

Early tests showed one AI tool detecting and mitigating 5,000 daily scam attempts to steal passwords, a feat Meta claims human teams could not achieve. Another AI reduced user reports of fake celebrity profiles by over 80% and doubled the detection of adult sexual solicitation content. The AI can also identify account takeovers by noticing suspicious activity like logins from new locations combined with password changes and detect fake websites spoofing legitimate businesses, the company says.

Source: The Register