发生了什么
BBC News Chinese reports that AI-generated videos featuring fictional medical experts are circulating among older viewers in Taiwan. The report describes a Taipei class where seniors used Gemini and ChatGPT to examine a video presented by a supposed doctor. Researchers and educators told the BBC that the videos are produced cheaply and at scale, often combining synthetic faces, voices and AI-assembled information. The BBC also found that the operators of many channels are not publicly identifiable. Claims about election-related manipulation remain unverified.
BBC News Chinese reports from a Taipei senior day-care centre where people aged roughly 60 to 90 learned to use phones and assess AI-generated videos. The article focuses on a featuring a supposed doctor called Zheng, wearing a white coat and stethoscope, discussing sarcopenia and claiming protein foods no longer work after age 60. One participant had received and forwarded it. In class, he asked Gemini and ChatGPT; the assistants reportedly said the doctor did not exist and that the video and health information had been generated or assembled by AI. The BBC does not independently provide a clinical assessment or establish the assistants’ accuracy.
The class teacher, Li Zi-qiang, told the BBC that AI-generated videos have circulated widely in Taiwanese seniors’ communities and messaging groups over the past year. The videos use virtual presenters claiming to be doctors, retired railway employees or former flight attendants. The Taiwan FactCheck Center told the BBC that content-farm YouTube channels produced false claims about health, senior benefits and transport discounts. Examples include fabricated claims that older passengers could automatically receive high-speed-rail business-class upgrades, free tickets for companions, cash gifts or free hospital caregivers. Taiwan’s rail operators have repeatedly said the purported senior benefits and hidden discounts do not exist.
Dai Yu-chen, a Taiwanese cybersecurity researcher, told the BBC he had tracked more than 200 YouTube channels carrying more than 70,000 videos with a combined 1.32 billion views. He estimated a video could be made in about 20 seconds using current tools: AI could create a doctor or expert’s face and background, synchronize an audio file to the mouth, assemble information from online material and add subtitles. These figures and estimates are the researcher’s claims reported by the BBC; the source does not describe an independent audit of the channel list, view counts or workflow.
The BBC says it examined 20 high-subscriber channels and found their operators disclosed no useful identifying information. Locations reportedly included Taiwan, the United States, Malaysia and Hong Kong, and some channels had been removed for violating platform rules. An anonymous Taiwanese national-security source alleged that some health and lifestyle channels changed into political advocacy during election periods. The allegation is not independently confirmed and does not establish a coordinated operation. China’s Taiwan Affairs Office has repeatedly denied allegations that China conducts influence operations against Taiwan and accused Taiwan’s government of manipulating public opinion.
为什么这很重要
The videos can make false health, welfare and transport claims appear authoritative, and the BBC reports that some viewers acted on those claims. The material also exploits trusted social relationships: older viewers may watch long videos for information or companionship and then forward them to friends. The scale described by a Taiwanese cybersecurity researcher suggests that synthetic misinformation can become a persistent part of the everyday information environment. The report does not establish that the channels are part of a coordinated political operation, but it documents why that possibility is receiving attention.
The immediate concern is not that every synthetic video causes measurable harm, but that false claims are presented through familiar signals of authority. The BBC reports that one class participant traveled to a high-speed-rail station because a video promised a benefit that did not materialize. The article reports no medical injury caused by the false doctor video, but health misinformation can influence choices about symptoms, treatment or diet. An authoritative-looking presenter and information tailored to older people’s concerns make the content harder to dismiss as obvious satire.
Forwarding is not simply information-sharing. An 86-year-old participant told the BBC she had trusted and forwarded material from friends and now felt deceived after learning that some videos were AI-generated. Another participant, spending several hours a day watching and sorting videos before forwarding them to at least 50 friends, described sharing as a way to maintain social connection. A Taipei psychiatrist told the BBC that health videos can respond to anxiety about aging by offering apparently actionable solutions, while forwarding can provide companionship and a sense of being useful.
Scale changes the problem. If the researcher’s estimates are accurate, a small number of channels can produce and distribute enormous volumes of plausible-looking material at very low cost. The BBC reports subjects including health, retirement, science, religion and welfare, topics closely tied to older viewers’ daily needs. YouTube’s parent company, Google, told the BBC its rules apply to AI-generated material and prohibit medical misinformation that could cause serious real-world harm. The report notes that many older viewers do not notice the AI labels attached to videos.
Political implications are significant but unresolved. A synthetic doctor or retired worker can build credibility around nonpolitical subjects before a channel introduces political claims. The BBC cites a 2024 study about lifestyle and health pages attracting Taiwanese audiences before political messaging was mixed in, but that background does not prove the channels here are coordinated. A digital-literacy expert told the BBC that even without malicious coordination, videos can function as AI slop: low-value or false material that consumes attention and makes reliable information harder to find. That documented information-quality problem is distinct from the unproven allegation of foreign political control.
互动机制：它实际上是如何运作的
以交互方式探索这一发展背后的基础技术。
crm_get_transaction(id='4092').Why can ethical evaluation not be reduced to one model score?
接下来看什么
The key questions are whether platforms can identify and limit high-volume synthetic misinformation, whether labels are visible and effective for older viewers, and whether independent investigators can verify the channels, view counts and production methods described to the BBC. Election-period monitoring will require evidence about coordination, ownership, funding and changes in content rather than anonymous allegations alone. Public-health professionals and media-literacy educators will also need to assess how people can verify medical claims without treating an AI assistant as a final authority.
Independent verification should first focus on basic empirical claims. Investigators would need to publish or document the list of channels, sampling methods, dates, view-count calculations and criteria for determining whether a video was generated by AI. The BBC reports one researcher’s estimates and its own review of 20 channels, but does not provide the underlying . It also does not establish how many viewers acted on medical claims, how often false videos are recommended by platform systems or how much content comes from a common operator.
Platform enforcement is another practical test. Google said the same rules apply to AI-generated videos and that serious medical misinformation is prohibited, while the BBC reports some channels were removed. Unknowns include how consistently labels are displayed, whether synthetic videos are demonetized, how recommendation systems treat them and whether removing one channel leads to rapid replacement by others. Those questions matter because the researcher described a network of channels with similar bot-account profiles, but the source does not independently establish one organization controls them.
Election-period monitoring should separate observable content changes from speculation. The anonymous national-security source cited by the BBC is not independently confirmed, so it should not be used as proof that channels shifted toward a particular political message. Stronger evidence would include transparent ownership links, repeated cross-channel coordination, common production assets, funding records or documented changes in distribution and messaging. The BBC reports that some channels have begun discussing whether unification with China is preferable, but the scale, origin and coordination of that content remain unknown.
The Taipei class offers a concrete response: viewers bring familiar videos to a lesson, question their claims and compare results with what they already know. That approach may improve skepticism, but the BBC does not establish that Gemini or ChatGPT can reliably authenticate a video or validate medical advice. Future coverage should track whether such classes change forwarding behavior, whether participants consult authoritative medical or government sources and how educators address the risk that one imperfect AI system is used to check another.