Jumio Study: Consumers Think They Can Spot Deepfakes Better Than They Can
Jumio surveyed 8,055 consumers in the UK, US, Singapore and Mexico: 52% believe they could detect a deepfake video, and 68% are open to a digital identity.
Identity verification company Jumio has published its Jumio 2023 Online Identity Study, the second edition of a yearly global survey of consumers carried out for it by Censuswide. The results show that people increasingly grasp how generative AI and deepfakes could speed up identity fraud, and why digital identities for online verification and authentication are needed. Yet many consumers also seem to rate their own ability to recognise a deepfake too highly, which may leave them more exposed.
Jumio describes itself as a leading provider of automated, end-to-end identity proofing, risk assessment and compliance solutions.
High awareness, misplaced confidence
Censuswide questioned 8,055 adults, divided equally between the United Kingdom, the United States, Singapore and Mexico, with fieldwork running from 4 to 6 April 2023. Censuswide follows the rules of, and employs members of, the Market Research Society, whose code is based on the ESOMAR principles.
More than two thirds of respondents (67%) knew of generative AI tools such as ChatGPT, DALL-E and Lensa AI, which can create fabricated video, images and audio. Singapore had the highest awareness at 87% and the UK the lowest at 56%.
Some 52% of those surveyed felt they would be able to identify a deepfake video. That points to overconfidence, since deepfakes have become sophisticated enough that the naked eye can no longer pick them out. The gap matters: the Singapore Police Force found that impersonation scams cost SGD101.3 million in 2022, a figure that may grow as the fakes get better. Jumio's own data also records a steady increase in the use of ever more sophisticated deepfakes worldwide and in many industries, with the payments and crypto sectors most affected.
Stuart Wells, Jumio's chief technology officer, observed that many people believe they can recognise a deepfake. There are giveaway signs, he said, but the technology is improving exponentially and is becoming ever harder to catch without help from AI. Businesses will increasingly need AI-driven tools to protect their networks and customers, he added, while individuals can defend themselves by approaching provocative images, video and audio with scepticism. "Some quick research will usually uncover whether it's a fake or not," he said.
Consumers see the risk to their identity
As familiarity with these tools grows, so does an understanding of how they might be turned to identity theft. Over half of respondents (57%) expect online identity theft to become easier because of them. Singapore consumers showed the strongest grasp of the potential for harm, at 73%, followed by Mexico (62%), the US (49%) and the UK (43%).
What Jumio says businesses should do
Philipp Pointner, Jumio's chief of digital identity, said organisations have a responsibility to teach customers about the subtleties of generative AI so they form a more realistic view of how well they can detect deepfakes. Even so, he noted, no amount of education will fully stop fraudsters from exploiting evolving technology. In his view, online businesses should adopt multimodal verification systems based on biometrics that can catch deepfakes and block the use of stolen personal data. He said the research showed strong consumer demand for this kind of verification and that companies should move quickly on it.
The survey found that 68% of consumers would be willing to use a digital identity to prove who they are online. The sectors where they would most prefer a digital identity to a physical document such as a driver's licence or passport were financial services (43%), government (38%) and healthcare (35%).