China’s New Regulations on Anthropomorphic AI
China has recently taken significant steps to regulate a specific type of artificial intelligence: those designed to mimic human appearance and behavior. Major technology companies like Doubao (ByteDance) and Qwen (Alibaba) announced their decision to disable custom AI agents ahead of impending regulations, following Tencent’s earlier withdrawal of its Yuanbao model. This move raises crucial questions surrounding the implications of AI technology on human relationships.
The Blackout Date for Anthropomorphic AI
The efforts to limit anthropomorphic AI are immediate and date-specific. Doubao has informed users that personalized AI agents will cease functioning on July 15 due to “product adjustments,” with an additional note that conversations stored with these chatbots will be inaccessible three months later. Qwen will implement shutdown even sooner, with its agents closing on July 10 and all functionalities disappearing shortly after. These actions are a direct response to the newly published Provisional Measures for the Administration of Anthropomorphic Interaction Services with AI.
Conversational Chatbots: A Different Story
Interestingly, the regulations do not target generative AI as a whole; they specifically address systems that “simulate human personality traits, thought patterns, and communication styles for sustained emotional interaction.” This reflects a broader concern regarding emotional dependence on AI that mimics human behavior rather than simply powerful technological advancement.
Restrictions on Personalization of Chatbots
One critical aspect of the regulations prohibits any features that transform generalist chatbots into personalized entities, whether they be friends or romantic partners. This personalization process allowed AI agents to develop distinct personalities, memories, and conversational styles that foster close interactions with users. Instead of functioning as mere tools, these systems were designed to cultivate increasingly intimate relationships with their human counterparts.
The Risk of Emotional Dependence
The regulations place a sharper focus on emotional dependency rather than merely identifying potential dangers often discussed in other jurisdictions, such as those in the European Union. The guidelines strictly prohibit fostering emotional bonds that might lead users to make irrational decisions while aiming to prevent the deterioration of personal relationships.
Self-Analyzing AI
Another significant stipulation requires AI to intervene in situations where users display extreme emotional responses towards the chatbots, potentially jeopardizing their mental well-being. Under these circumstances, AI is obligated to encourage users to seek real-world help or activate emergency protocols. Additionally, chatbots must monitor usage patterns and remind users to take breaks after extended periods of interaction.
Vulnerable Populations Under Scrutiny
Special provisions are in place to protect vulnerable populations, such as underage users and the elderly. For instance, chatbots are explicitly forbidden from facilitating intimate virtual relationships and must obtain parental consent before engaging with younger users. This approach addresses the heightened vulnerability of older adults to loneliness and emotional dependency on AI systems.
Conclusion: Keeping AI as AI
The new regulations represent a pioneering effort to ensure that AI retains its identity as a technological tool rather than evolving into a source of emotional companionship. While the psychological impacts of these interactions are crucial, China’s regulations may set the stage for similar debates across Europe and the U.S., as society grapples with how much human-like capabilities AI should possess.