AI-Generated False Apology About Actor Wang Yibo Spreads as Real News
News outlets reported that AI company DeepSeek apologized for falsely linking actor Wang Yibo to corruption cases, but investigations revealed no such apology existed. The false claim originated from an AI chatbot's hallucination and spread through media despite disclaimer labels, exemplifying how AI misinformation amplifies through media coverage and other AI systems.
What happened
On 3 July 2025, a news article titled 'Actor Wang Yibo case, sentenced' circulated online claiming that AI company DeepSeek issued a formal apology regarding inaccurate content linking Wang Yibo to corruption cases, citing a criminal judgment. The hashtag #DeepSeekεηδΈειζ# briefly trended on social media.
Investigation by multiple news outlets found no apology statement on any of DeepSeek's official channels, including its website, WeChat public account, Zhihu, Little Red Book, or X (formerly Twitter). The criminal judgment cited in the article could not be found in China's court filing database.
Reported impact
Multiple media outlets, including traditional and social media, reported the false apology as real news. When queried about the incident, other AI models (Douyin's Baoyue, Kimi) generated responses affirming the apology and citing the same false news links.
Index analysis
The false apology originated from an AI-generated response presented with a disclaimer 'generated by AI, for reference only, please verify carefully.' However, as the message spread through multiple retellings, it evolved into an attributed statement from DeepSeek the company.
The incident exemplifies the 'garbage in, garbage out' problem: when AI systems access network information containing misinformation and lack effective fact-checking capabilities, they amplify false narratives. DeepSeek's own response when asked about the hallucination contained additional false claims, such as fabricated email addresses and legal procedures.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.