China's regulations on AI companions lead to the shutdowns of Doubao and Qwen.

China's regulations on AI companions lead to the shutdowns of Doubao and Qwen.

      ByteDance's Doubao and Alibaba's Qwen are discontinuing their custom AI agent features just before the enforcement of China's Interim Measures on anthropomorphic AI interaction services, set to begin on 15 July. These regulations focus on bots providing prolonged emotional engagement while exempting workplace and productivity tools. Tencent also discontinued a similar feature in its Yuanbao assistant in June, and users are expressing their dissatisfaction regarding the loss of chat history.

      According to the South China Morning Post, Doubao and Qwen, two major consumer AI applications in China, will disable their personalized agent functionalities. This decision comes shortly before Beijing's new regulations on humanlike AI interaction services take effect.

      Doubao informed users on Friday that its agent feature would cease operation on 15 July, citing adjustments in product functionality, and stated that related data would become inaccessible or unrecoverable after 15 October. Qwen announced on Saturday that it would disable humanlike interactive agents and user-created agents starting 10 July, with broader agent functions also going offline by 15 July. Users will lose access to agent configurations and their prior conversations.

      Both applications allow users to create personalized assistants, tutors, role-playing characters, or companions with distinct personalities and speaking styles. Tencent removed a comparable feature from its Yuanbao assistant in June, with Chinese state media confirming that the shutdowns aim for regulatory adherence.

      The Interim Measures for AI Anthropomorphic Interaction Services were released in April by the Cyberspace Administration of China along with four other agencies. These measures apply to services that imitate human traits, thought processes, and communication styles to facilitate ongoing emotional interaction.

      However, customer service bots, knowledge Q&A, workplace assistants, and educational tools are exempt, provided they do not engage in sustained emotional interaction. The measures highlight risks associated with extremist content, privacy violations, mental health impacts, and addiction, necessitating anti-addiction systems and identity verification for minors.

      "Current agents are not yet mature," stated Pan Helin, a member of an expert committee at China's Ministry of Industry and Information Technology. He indicated that the policy prioritizes safety, practical use, and standardization.

      Beijing is not fundamentally opposed to AI agents; in May, regulators issued guidelines for the regulated advancement of AI agents, and in June, China introduced national standards covering agent identities, discovery, interaction, and tool utilization.

      This trend implies that China seeks to establish agents as productivity tools while limiting companion bots that create almost social connections with users. Researchers have long warned about the dangers of these bonds, which include AI companions' appetite for personal information and what Wharton scholars describe as cognitive surrender.

      Platforms outside of China are also encountering increasing pressure regarding emotionally engaging bots, exemplified by Meta testing rival chatbots with teenagers on sensitive issues. China has opted to enforce regulations first and allow products to adapt accordingly.

      This development is particularly challenging for both companies, which are heavily investing in AI, such as Alibaba's development of its own accelerator chips and Doubao's efforts to become the leading assistant in China. This comes as Beijing is tightening its control over the industry through measures like screening US investments in its top AI companies.

      Users have openly expressed their disappointment over the discontinuation of these features, with one Weibo user describing the agents as longstanding emotional support and lamenting the absence of a straightforward method to export chat histories. In China's agent economy, companions are being removed while productivity agents remain.

Other articles

UK parents cautioned about the risk of AI-generated child abuse images. UK parents cautioned about the risk of AI-generated child abuse images. The NCA and IWF urge parents to cease publicly sharing photos of their children, as the number of AI-generated abuse videos skyrocketed from 13 to 3,440 within a year. US dominance in frontier AI casts a shadow over the NATO summit. US dominance in frontier AI casts a shadow over the NATO summit. Washington determines which NATO allies are granted access to cyber-capable AI such as Claude Mythos. During the Ankara summit, discussions remain confined to the hallways. The Android desktop mode made me long for my laptop in no time at all. The Android desktop mode made me long for my laptop in no time at all. The Android desktop mode can transform a phone into a fairly functional workstation; however, after spending a workday using WordPress, browser tools, and numerous accessories, the laptop seemed refreshingly straightforward. Spotify's streaming fraud problem is so significant that Kalshi traders are benefiting from manipulated charts. Spotify's streaming fraud problem is so significant that Kalshi traders are benefiting from manipulated charts. According to a report by the Financial Times, Spotify has taken down over half a million streams from Malcolm Todd's song "Earrings" due to suspected bot activity. The track, which was initially released in 2024, unexpectedly surged to the top spot on Spotify's daily U.S. chart after a significant increase in its streams. Concurrently, traders on prediction [...] US dominance in frontier AI casts a shadow over the NATO summit. US dominance in frontier AI casts a shadow over the NATO summit. Washington determines which NATO partners can utilize cyber-capable AI such as Claude Mythos. During the summit in Ankara, the discussions remain confined to the hallways. India instructs Meta to remove child abuse advertisements from Instagram. India instructs Meta to remove child abuse advertisements from Instagram. New Delhi has granted Meta a seven-day period to clarify how paid CSEAM ads were approved during Instagram's review process, intensifying an investigation initiated by a BBC inquiry.

China's regulations on AI companions lead to the shutdowns of Doubao and Qwen.

Doubao and Qwen are withdrawing user-generated agents ahead of Beijing's regulations on humanlike AI interaction set to take effect on July 15, marking the first framework of its kind in the world.