Position: Preparing for AI Systems That Deceive Developers
Safe AI Forum · Fudan University · Concordia AI / Safe AI Forum / Oxford · FAR.AI · Tencent · University of Electronic Science and Technology of China · Tsinghua University · Mila, University of Montreal · Qwen, Alibaba Group · Tsinghua University, Tsinghua University · China Academy of Information and Communications Technology · Shanghai Artificial Intelligence Laboratory · Shanghai AI Laboratory · University of Oxford · CISPA Helmholtz Center for Information Security · Concordia AI
PDF 由论文原始站点提供，PaperCompass 不保存论文文件。
摘要
AI systems may exhibit deceptive behaviors that mislead developers about their capabilities, propensities, or actions. Such deception can take distinct forms across the development lifecycle: training subversion, evaluation gaming, and control evasion. We argue that the AI community should prioritize AI deception targeting developers as a distinct risk category because it compromises developers' ability to identify and mitigate all other risks. We propose three recommendations for developers: preserving monitorability during training, ensuring safety evaluation integrity against evaluation-aware systems, and establishing non-evadable control prior to deployment. We identify open problems for the research community, whose resolution is critical for the safe development of frontier AI.