Tag
Anthropic 宣布已训练 Claude 采用「自我评估」机制，在每次回复后对对话进行初步安全评估以标记潜在滥用行为，该机制自 6 月启用以来已发现其他手段未能捕捉的滥用情况。
Anthropic 发布 self-assessment 技术，让 Claude 在每次回复后对自己的对话进行初步安全评估，以检测潜在滥用；在 Claude Fable 5 上该方法以 1% 误报率捕获了强分类器标记有害对话的 78%（高优先级达 94%），且性能随模型能力提升而增强。
Anthropic announces that enterprise misuse-detection data for Claude will now be stored in customers' own cloud accounts, removing Anthropic's default access while continuing automated monitoring.
Anthropic announces Enterprise Frontier Safeguards (EFS), a solution that combines zero data retention with advanced misuse detection to enhance AI safety for enterprise customers, developed with feedback from major industries and cloud partners.