Taiwanese adults have embraced artificial intelligence at a rate that outpaces most of the world — but a new survey suggests their ability to recognize when AI is being turned against them has not kept pace.
Research published Tuesday by TrendLife, the consumer division of cybersecurity firm Trend Micro, found that 74.9% of Taiwanese adults have used AI tools to help navigate significant personal decisions, from job searches and investment choices to government benefit applications and major purchases. The figure is nearly double the 46% recorded in comparable global surveys, signaling that Taiwan has moved well beyond early AI curiosity into routine reliance.
Yet the same survey of 1,035 adults revealed a striking vulnerability: only about 26% of respondents said they were "very confident" or "extremely confident" in their ability to spot AI-generated fraud or deepfake content. Just 5% selected the highest confidence level.
High Adoption, Low Detection Confidence
The tools Taiwanese consumers report using include ChatGPT, Claude, Google Gemini, Amazon Rufus and Microsoft Copilot. Their applications have moved well beyond casual information-gathering. Respondents described using AI to weigh options, draft communications and support decisions in high-stakes scenarios — precisely the contexts that scammers increasingly target.
Trend Micro said this behavioral shift has created a new attack surface. As users grow accustomed to acting on AI-generated guidance, fraudulent content engineered to mimic that same voice and format becomes harder to dismiss on instinct alone.
Basic Digital Hygiene Is in Place — but AI Fraud Plays in a Grey Zone
The survey results do not suggest Taiwanese internet users are careless. Among respondents who had shared personal data online, 67.4% said they use two-factor or multi-step authentication, and 67.3% said they actively monitor bank and account activity for irregularities. Around 53.5% said they avoid public Wi-Fi, and 52.3% said they verify the legitimacy of websites or institutions before engaging.
The problem, Trend Micro argued, is that these habits were built for an earlier threat landscape. Previously, fraud was often detectable through obvious tells: grammatically awkward text, low-resolution images, suspicious URLs. Generative AI has eroded those signals. Scam content today can be grammatically polished, visually coherent and situationally plausible — designed to look like legitimate investment advice, a government benefits notice or an endorsement from a trusted figure.
"The most dangerous traps are no longer obviously suspicious," the company said in its findings. "They are scenarios that appear reasonable but already carry concealed red flags."
Social Video Becomes a Primary Entry Point for Deepfake Fraud
Social media platforms have emerged as the leading vector for AI-driven scams, the survey indicated. Fraudsters are increasingly deploying deepfake videos impersonating celebrities, AI face-swap content and short-form clips on platforms including Facebook, X and TikTok — then redirecting users toward messaging apps, private investment groups or counterfeit websites.
This multi-step funnel is what makes the new generation of scams particularly effective. By the time a user thinks to verify a URL or scrutinize the wording of a message, they may already have been moved several steps deeper into the scheme. Traditional detection habits, Trend Micro noted, apply too late in the process to reliably interrupt it.
Nearly 40% of Taiwanese Are Already Using AI to Fight Back
One of the more striking findings from the survey is that consumers are beginning to deploy AI defensively — not only as a decision-making aid, but as a fraud-screening tool. Some 38.9% of respondents said they had already used an AI tool to assess whether online content might be fraudulent.
Trust in those tools, however, remains measured. Among people who had used AI for scam detection, only 5% said they completely trust AI-generated risk assessments. A larger share — 39.0% — said they mostly trust them, while 46.4% said they trust them to some degree. Trend Micro interpreted the results as evidence of genuine consumer demand for AI-assisted fraud detection, paired with a clear expectation that the tools must do more than return a verdict. Users want risk assessments that are clear, understandable and verifiable, particularly in ambiguous situations where the threat is not self-evident.
Trend Micro Expands Detection Capabilities to Social Video
In response to the evolving fraud environment, Trend Micro said its AI-powered scam detection product has been updated with high-risk category classification, designed to give users more granular alerts when they encounter suspicious content. The company has also extended the tool's scanning capabilities to cover scam video links shared on Facebook, X and TikTok — expanding beyond text messages and URLs.
The company also recommended that households develop coordinated fraud-awareness practices, so that family members can consult one another before responding to suspicious messages, investment solicitations, wire transfer requests or demands for personal data. As AI tools become embedded in daily routines, Trend Micro said, scam-detection capabilities must advance in parallel — adding a layer of scrutiny before each click, transaction or data submission, rather than attempting to recover after the fact. （Related： The End of the Subscription Model: How AI Agents Are Reshaping the SaaS Economy ｜ Latest ）
