深度共情?陪伴型AI要厘清“人机边界”

Core Viewpoint - The release of the "Interim Measures for the Management of Human-like Interactive Services of Artificial Intelligence (Draft for Comments)" marks a new phase in the governance of AI human-like interactive services in China, focusing on systematic and precise regulation to balance innovation and risk prevention [7][8]. Regulation and Management - The draft defines AI human-like interactive services as those that simulate human personality traits and communication styles, emphasizing the need for significant user notifications that they are interacting with AI rather than a human [3]. - It establishes eight prohibited activities, including generating content that harms national security or spreads misinformation, and mandates service providers to implement safety responsibilities throughout the service lifecycle [4]. User Protection and Data Security - The draft emphasizes the protection of vulnerable groups, particularly minors and the elderly, by requiring features like a minor mode, consent from guardians, and emergency contact settings [5]. - It mandates that training data must align with socialist core values, ensuring data legality and traceability, and requires user consent for the use of interaction data in model training [5]. Safety Assessment and Reporting Obligations - Service providers are required to conduct safety assessments and report findings when user numbers reach certain thresholds or when significant changes occur, promoting a self-regulatory mechanism [6]. - A tiered regulatory approach is proposed for addressing violations, including warnings and service suspensions [6]. Expert Opinions - Experts highlight that the draft aims to create a dual governance model that emphasizes both positive guidance and risk prevention, particularly focusing on the emotional interaction risks associated with AI [7]. - The core philosophy of "responsible innovation" is emphasized, with a focus on clarifying the boundaries between humans and machines [7].