A controversial new application of artificial intelligence is sweeping across Chinese social media, allowing users to resurrect past relationships in digital form. The South China Morning post reports that by uploading chat logs, social media posts, and photographs, individuals are creating highly personalized AI replicas of their ex-partners, igniting a fierce national debate about emotional fidelity, data privacy, and the boundaries of digital attachment.
The Technology Behind the Trend
The trend originated from an open-source project called “Colleague.skill,” developed by Shanghai-based AI engineer Zhou Tianyi. Initially designed to generate reusable AI skill packages from workplace communications, allowing a user to create an AI that could respond in the style of a colleague, the technology was quickly adapted by developers into “Ex-partner.skill.” This package allows users to train a virtual ex that can engage in conversations using familiar tones, thought patterns, and shared memories.
The process requires significant effort and intimate knowledge. Users refine the replica by feeding it details about past trips, eating habits, anniversaries, and even past arguments. Some users have successfully integrated these digital avatars into messaging apps like WeChat, effectively creating a persistent digital companion. Due to the technical complexity involved, some netizens have paid between 25 and 45 yuan ($3.70 to $6.60) for installation services from third-party developers.
One user, posting under the handle @Bingtangcheng, described spending an entire night training the AI. “When his ex-girlfriend finally ‘appeared’ on WeChat, he exclaimed: ‘All my efforts were worth it,'” the South China Morning Post reported. The phenomenon also extended to public figures; some programmers uploaded biographies to create digital replicas of celebrities like Elon Musk and Steve Jobs.
The Wave of “Cyber Heartbreak”
As older AI models were phased out in March 2026, a wave of “cyber heartbreak” spread across mainland social media as users lost access to their carefully crafted digital companions. This phenomenon prompted psychologists and legal experts to weigh in on the profound implications of the technology for human relationships and the law.
Wanqiu, a marriage consultant based in Guangdong, suggested that creating an AI ex is a “normal emotional reaction” to grief. “As long as it does not harm your current partner, it shouldn’t be seen as cheating,” she stated. However, she cautioned that these digital replicas could foster unhealthy emotional attachments, potentially hindering individuals from forming genuine relationships in the real world. The AI, by design, is infinitely patient and agreeable, a distorted mirror of the real person, stripped of the friction that defines authentic human connection.
Privacy and Legal Implications
The legal implications are equally complex. Zhong, a Guangdong-based lawyer, warned that accessing and utilizing an ex-partner’s chat records and personal data without their explicit consent could constitute a violation of China’s Personal Information Protection Law. The unauthorized creation of a digital likeness raises profound questions about digital consent and post-relationship privacy that existing legal frameworks are ill-equipped to address.
Beyond the legal and psychological concerns, the trend has sparked philosophical reflections among Chinese netizens. While some find solace in the digital echoes of past loves, others argue that the technology fundamentally misunderstands human connection. As one netizen poignantly observed, “The frustration, confusion, happiness, and anger one experiences in life cannot be digitized. This is the irreplaceable essence of humanity.” The debate ultimately reflects a broader societal reckoning with the power of AI to simulate intimacy and the question of whether simulation is the same as the real thing.
The Regulatory and Platform Response
The trend is unfolding against a backdrop of rapidly evolving AI governance in China. The Cyberspace Administration of China’s draft rules on AI-generated content, which require clear labeling of AI-generated material and prohibit the unauthorized use of a real person’s likeness, would, if applied strictly, render the Ex-partner.skill application illegal in its current form. The rules have not yet been finalized, but their direction is clear: the CAC views the unauthorized digital replication of real individuals as a category of harm that requires explicit regulatory intervention.
Platform operators are also beginning to respond. WeChat’s parent company Tencent has not yet issued a formal policy on AI-generated digital companions integrated into its messaging platform, but the company’s existing terms of service prohibit the use of automated accounts to simulate human users, a provision that could be interpreted to cover AI replicas operating within the WeChat ecosystem. The fact that users are paying third-party developers to install these systems suggests the practice is already operating in a regulatory gray zone unlikely to remain unaddressed for long.
The Ex-partner.skill phenomenon is, in many ways, a preview of the broader regulatory challenges that China’s AI governance framework will need to confront as generative AI becomes more capable and more personal. The NRTA’s first major AI content labeling crackdown earlier this year established the principle that AI-generated content must be identifiable as such. Extending that principle to AI-generated personas, digital replicas of real people, operating in real communication channels, is the next frontier.