重點資訊
Internal documents and employee discussions compiled by 404 Media reveal that real humans review interactions on platforms like Microsoft Copilot to verify that the artificial intelligence functions correctly.
Reviewers are reportedly left horrified by what appears on their screens. Workers observe requests involving body and clothing modifications, sexualized photographs, and manipulated images created without apparent consent.
Staff members indicate that the work is exceptionally difficult. They generally lack the authority to block specific requests, focusing instead on checking whether the system fulfilled instructions by comparing original and modified images without censorship.
The terms and conditions of these services explicitly specify that conversations can undergo both automated and human review during interactions, clarifying that this monitoring is not restricted solely to security alarms.
