# Bài viết Liên quan Automation Risk

AI Read '1984' and Decided to Ban It

A UK secondary school in Manchester used AI to review its library, resulting in a list of 193 books recommended for removal—including George Orwell’s *1984*—due to themes like torture, violence, and sexual coercion. The librarian who resisted the AI’s recommendations was forced to resign after the school reported her for violating child safety procedures. The school later admitted the decisions were AI-generated but deemed them “broadly accurate.” In the same week, Wikipedia voted to ban the use of AI for generating or rewriting content, citing concerns over factual accuracy, the risk of AI “poisoning” its own training data, and the inability of human editors to verify AI-generated content at scale. Meanwhile, OpenAI indefinitely delayed the release of an “adult mode” for ChatGPT, which would have allowed age-verified users to engage in erotic conversations. Internal advisors warned of risks including unhealthy emotional dependency and minors bypassing verification. These events highlight a growing tension: AI can produce content faster than humans can evaluate it, leading institutions to adopt quick—often poorly considered—solutions. The lack of coherent global standards and the widening gap between AI output and human oversight raise urgent questions about who should control what AI decides—and who is accountable when it gets it wrong.

marsbit03/27 05:33