Meta引入一个新的AI框架,对高风险AI系统进行分类和控制,以防止滥用。 Meta introduces a new AI framework to categorize and control high-risk AI systems to prevent misuse.

Meta发布了一个“前一级AI框架”,根据AI系统在网络安全、化学和生物攻击方面的潜在帮助作用,将其分为高度和关键风险类别。 Meta has released a "Frontier AI Framework" that classifies AI systems into high and critical risk categories based on their potential to aid in cybersecurity, chemical, and biological attacks. 高风险系统获得和减轻风险的机会有限,而关键风险系统将停止发展,直到更安全。 High-risk systems will have limited access and mitigations, while critical-risk systems will halt development until safer. 该框架旨在防止灾难性后果,尽管Meta承认保护措施的局限性。 The framework aims to prevent catastrophic outcomes, though Meta acknowledges limitations in protection measures.