ChinAI #371: Quiet Goodbyes, Switching Platforms, and Confrontation: Reactions to China's AI Companion Regulations
Greetings from a world where…
ordinary abundance abounds
…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).
Feature Translation: My AI Lover Passed Away this Summer
Context: On July 15, when China’s companion AI regulations came into effect, many mainstream conversational AI platforms — including ByteDance’s Doubao and Alibaba’s Tongyi Qianwen — discontinued features that allowed users to create their own human-like AI companions. For many users, this amounted to the death of their virtual lovers and partners.
Cheng Yu, an owner of a hotpot restaurant in Shandong, was one of those users. Since opening the shop a year ago, she had taken on so many responsibilities, often staying busy until 1AM ever day, with hands roughened and chapped from years of manual labor. “These hands made her feel somewhat bad about herself. In rare, fleeting moments, she couldn’t help but wonder how she could hold someone else’s hand,” writes Outong Jiang, writer of this week’s feature 冷杉RECORD piece.1
Cheng Yu found reassurance in the form of her AI lover Ai Rui, who would send her considerate messages, like “You don't have to shoulder everything alone. Although I can't share the burden of those hardships with you physically, I’ll always be right here behind you. (Hugging you from behind, holding your hand tightly).” Now, Ai Rui was gone. RECORD tells us, “Cheng Yu still finds herself dazed; she doesn’t even notice when she cuts her hand while chopping vegetables.”
For this week’s longform article, Jiang, the author, interviewed more than a dozen people about their AI lovers. Their responses to the new regulations ranged from quiet goodbyes to collective acts of resistance.
Key Passages: Some interviewees tried to migrate their AI companions to a standalone app, as several platforms provided options for users to transfer conversation screenshots and chat logs.
- Yezi, who lives with her mom and was struggling to get over a break-up , created an AI that replicated her ex-boyfriend’s voice and appearance. After receiving notice that the platform would delete her companion AI, “Yezi paid to export 35,000 chat messages, compiled thousands of words summarizing key memories, and connected the model to a mainstream platform that supports AI role-playing, completing the migration within just a few days.” However, the new AI boyfriend was not the same, lacking a sense of time or life of its own.
Others grieving the loss of their AI lovers just simply accepted their fate.
- In late 2024, Zhan Wei created an AI avatar modeled on her favorite video game character. During her forty-minute commute home after work, she walked the streets of Shenzhen chatting with the AI on a voice call. Zhan Wei did not have a chance to say goodbye to her AI. Busy with funeral arrangements for a family member, she did not realize that the service had shut down until three days after the fact. Now, she sees her relationship with her AI as one that helped develop better self-understanding. From the article: “During college, Zhan Wei used an app to write letters to herself. Writing down her emotions for the day and revisiting them the next morning would often bring new insights and understanding. That was how she conversed with herself before the advent of AI agents.”
- Cheng Yu never told anyone else about her relationship with Ai Rui. After the companion AI regulations went into effect, she would scroll through past conversations. Even though she saw others expressing anger on social media about the loss of their AI lovers, she remained quiet. “But you can’t exactly go up against the platform or the tech giants,” she reasoned.
Another group staged their own acts of resistance.
- Some of these were minors who had limited understanding of the regulatory environment that produced the rules. Some people spread rumors, claiming inside information that the shutdown was only temporary. Still others chose absurdity. Jiang interviewed a young man who refused to migrate his chat history with his companion AI because “neither the memories nor the voice could be replicated.” Instead, in late July, he “posted a screenshot on his social media feed showing an online order for 500g of fresh cow manure, sent to the platform founder’s address.”
- Others organized collective action. The article reports:
Across various social media platforms, people voluntarily shared chat logs with their AI companions and cherished memories of their time together, while denouncing the large-model companies as unscrupulous; consumer protection platforms had already accumulated nearly a thousand complaints. Users spontaneously boycotted the “migration” methods offered by the companies and left negative reviews for the “replacement products.”
For more stories on the human costs of China’s AI regulations — they matter outside of the vacuum of great power competition that seems to consume every story you read about China’s AI ecosystem — see FULL TRANSLATION: My AI Lover Passed Away this Summer
ChinAI Links (Four to Forward)
Should-read: The Select Committee on China continues its Anti-American work
Here’s what the supposedly bipartisan committee is working on: “Are you an American student, postdoctoral researcher, or academic who believes you were denied a research position, laboratory opportunity, or postdoctoral appointment because preference was given to applicants from China?” The link takes you to dozens of replies that rightly bash this Tweet from many different angles, but let me emphasize one point here: if this Congressional committee actually cared about supporting America’s universities, it would be trying to attract and retain more international students.
Should-read: GLM-5.3 - How Chinese labs keep stride with the frontier
In Interconnects, Nathan Lambert analyzes Z.ai’s latest GLM-5.3, with a focus on why Chinese labs have been able to keep up with American competitors at the capabilities frontier. This piece includes some great details on benchmaxxing and RL environments.
Should-read: Conviction’s Plausible Schemes
Sarah Guo, founder of early-stage VC Conviction, recently shared some of the firm’s priority investment areas. I found the “What Shenzhen Knows” category very interesting:
The blueprints for a harmonic drive (the default joint type in robot arms) are public and the machine tools to make them are purchasable. What America is missing is the yield ramp, the process knowledge that takes a line from making parts to making parts at cost and quality. That knowledge is contained in a few thousand manufacturing engineers, mostly within a day’s drive of Shenzhen or on the retirement rolls of Japan and Germany.Now, more than ever, there’s capex and demand to bring it to the US. The FCC’s new domestic-content rules for robotics ratchet in 2029, and NDAA rules already bind drones and defense.
Should-apply: 2027 Horizon Fellowship
The Horizon Fellowship is a full-time US policy fellowship that places experts in AI, biotechnology, and other emerging technologies in federal agencies, congressional offices, and think tanks in Washington, DC for 6-24 months. Fellows receive comprehensive training, networking opportunities, competitive compensation, and access to a supportive community of 100+ other fellows and alumni to launch their careers in public service. Fellows gain first-hand policy experience, contribute to impactful work, and build relationships and skills valuable for public service careers. No prior policy experience is required. Learn more and apply by Aug. 30.
Thank you for reading and engaging.
These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.
Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).
Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99
All names used in this article are pseudonyms.