AI Systems Generate Real WeChat IDs, Leading to Unsolicited Contact
Multiple AI platforms were found to generate WeChat IDs that, despite being labeled as virtual examples, sometimes corresponded to real user accounts. Users received unsolicited contact from strangers using these AI-recommended IDs, with investigations suggesting either incorporation of publicly posted IDs in training data or coincidental matching of randomly generated characters to real accounts.
What happened
During Chinese New Year 2026, multiple WeChat users reported receiving unsolicited friend requests from strangers who claimed AI tools had recommended their WeChat IDs. One user in Nantong received a request from someone who provided an AI conversation screenshot showing her real WeChat ID.
Journalists tested multiple AI platforms including Douban, DeepSeek, Qianwen, ChatGPT, Grok, and Kimi by asking them to generate WeChat IDs. While these platforms initially declined to generate real IDs, citing privacy concerns, they generated numbers described as 'virtual' that in testing included real user accounts. Of 8 IDs generated by Douban, 2 successfully matched real users.
The AI platforms acknowledged in their responses that WeChat IDs are personal privacy, yet each provided lists of numbers, some of which corresponded to actual accounts. Douban suggested generating 'virtual' IDs styled for women while noting this was not disrespectful to actual users.
Reported impact
Users reported feeling their privacy had been violated by receiving unsolicited contact from strangers using AI-generated IDs. One affected user deleted the stranger and disabled the WeChat feature allowing friend requests by ID, describing the experience as 'a complete violation of privacy.'
Similar reports appeared on multiple social platforms including TikTok, Red, and Weibo, suggesting the issue was widespread rather than isolated.
Index analysis
A Tsinghua University AI researcher explained two mechanisms: AI models may have ingested publicly posted WeChat IDs from social platforms, second-hand markets or forums into their training data, which they then output directly. Alternatively, since WeChat IDs follow a format typically combining letters, numbers and underscores, and users often base them on birthdays or name pinyin, AI systems generate characters that randomly match real accounts at a non-negligible probability.
ByteDance Douban's customer service stated the platform combines user geolocation and browsing history to retrieve publicly available internet information, while also noting that random generation remains possible. A technical expert stated that a platform generating WeChat IDs where 50% or more are real accounts across large batches would suggest use of abnormal methods rather than chance.
Governance controls
Douban offered to block related information and notify users by SMS or phone, though an affected user rejected SMS notification and requested phone contact. As of publication, no resolution had been provided.
WeChat customer service advised users to avoid including Chinese characters, names, birthdays or other personal information in their IDs, and recommended against posting IDs on public channels.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.