How Did a Deepfake Scam Trick Hong Kong Out of Millions?
PLUS: ByteDance’s China CEO Resigns; SMIC Profit Drops 55%; TSMC Japan strategy
By: JINPENG LI
When synthetic media were used to directly influence elections across the internet and threaten democracy in Western societies, the aim of deepfake crimes in Asia was more straightforward: to make money. Today, let’s talk about the large-scale deepfake finance fraud in Hong Kong.
More targeted and expensive
The fraud took place within a multinational company, where a spammer acquired audio and video clips of individuals from the company through YouTube and other media sources, then used them to recreate a synthetic video conference call.
The employee mentioned that this multi-person video conference seemed incredibly realistic, with colleagues looking and sounding like real people he recognized. Even the CFO attended. Thus, he put aside his initial doubts and followed the instructions, executing 15 ‘confidential transactions’ totaling HK$200 million-approximately $25.6 million-to five Hong Kong bank accounts.
According to HK media, the fraudulent conference was a pre-recorded video where fake superiors issued orders, preventing any real interaction between the employee and the supposed participants. The employee asked only for a brief self-intro. Post-conference, the fraudsters kept in contact via instant messaging, emails, and direct video calls
In this HK$200 million case, there are many doubts, but notably, Deepfake scams in China are becoming more targeted and costly as technology grows more advanced and accessible. Scammers carefully choose their victims, needing detailed knowledge of their roles, reporting lines, and financial authority. They craft complex videos, even collecting information on normal colleagues for authenticity.
Another concerning Deepfake scams in China involves cyber kidnapping, primarily targeting families of Chinese students studying abroad. Criminals deceive families into believing their child has been kidnapped by playing screams synthetically generated to mimic their children’s voices, even though the child is safe.
Additionally, these scammers use social media to research and gather information on targeted children before making contact, enabling them to create more realistic voices or images.
China’s regulation
In late 2022, China issued the world’s first deep fake regulation, which requires that material have the subject’s consent and bear digital signatures or watermarks, and the Deepfake service providers must offer ways to “refute rumors”. The state-affiliated media, People’s Daily, specifically emphasized that algorithm models like OpenAI must also comply with these regulations in China.
Tiffany Hsu at New York Times:
But China faces the same hurdles that have stymied other efforts to govern deepfakes: The worst abusers of the technology tend to be the hardest to catch, operating anonymously, adapting quickly and sharing their synthetic creations through borderless online platforms.
However, Hsu might have underestimated the Chinese government’s regulatory capabilities. According to interviews with experts by the People’s Daily, China plans to focus the next phase of deepfake regulation on building a monitoring and management platform for the mobile internet and all media types. This will also involve strengthening a hybrid regulatory approach that combines technology and human moderation. In my imagination, this means that if a deepfake that disrupts social stability goes viral, such platforms would quickly detect and effectively ban it.
More Big Techs Join C2PA to Address the Deepfake Issues
Meta, Google, and OpenAI all announced their new measures regarding AI-generated content last week, which include visible labels and invisible watermarks in images or videos, following the Taylor Swift Deepfake incident. Both three Big Tech are joining the C2PA, the Coalition for Content Provenance and Authenticity, to detect the sources of AI-generated content.
C2PA was launched two years ago as an open-source internet protocol. It uses cryptography to embed information about the origins of content-what technologists call “provenance” information. The organization also includes participation from other companies such as Microsoft and Adobe.
Asia Must Reads
ByteDance’s China CEO Resigns to Focus on AI
Kelly Zhang has announced she will join the team at Jianying, a Chinese editing app akin to CapCut, which is owned by ByteDance. She aims to contribute to the app’s growth in the era of AI.
Why it matters: ByteDance often initiates multiple projects to compete with each other, setting a clear strategic direction. According to LatePost, ByteDance currently has several teams working on AI innovation, focusing on three major models in visual and textual domains, with applications developed by teams codenamed Flow, Ocean, and others. However, ByteDance has not been satisfied with the outcomes of these AI projects. Kelly Zhang’s move to Jianying signals an increase in ByteDance executives committed to AI application innovation.
SMIC Profit Drops 55%
China’s largest chipmaker, SMIC, reported a significant drop in its fourth-quarter net profit late Tuesday, falling 55% year-on-year to US$174.7 million, despite a 3.5% increase in revenue, due to rising costs.
Why it matters: SMIC’s results dampen hopes for a quick recovery in the industry, citing fierce competition and a period of high investment. It highlighted that the chip industry is at a low point in its cycle. From a broader market perspective, the strength of demand recovery is not yet sufficient to support a full rebound in the semiconductor industry, with capacity utilization rates unlikely to return to high levels in the short term.
Bonus
Semiconductor manufacturing giant TSMC is expanding its production capacity in Japan. TSMC announced it would further invest in its Kumamoto-based wafer manufacturing subsidiary, JASM, to build a second wafer plant. Construction of the new wafer plant is set to begin at the end of 2024, with operations expected to start by the end of 2027. ( Jane Lee / Bloomberg)
Taiwan’s PC makers Acer and Asutek plan to step up manufacturing in India this year as the country weighs tightening import restrictions to boost the local tech industry. ( Lauly Li / Nikkei Asia)
Wenxin Yiyan launches new digital avatar feature. Users only need a photo and to record three sentences to create their exclusive digital avatar, which also supports personalized definitions of names, voices, MBTI personalities, and more.(Jiemian)
China and Russia have agreed to consult and coordinate on the military use of artificial intelligence (AI). Meanwhile, a year ago, Washington initiated the Political Declaration on the Responsible Military Use of AI and Autonomy, aimed at establishing international norms. Although 46 countries have endorsed it, China and Russia have not signed on. ( SCMP)
China use the new chipmaking approach that breaks down chip into independent modules to reduce design costs and improve computing performance to develop more powerful chips despite the US government sanction.（ Zeyi Yang / MIT Tech review)
Originally published at https://theabacus.substack.com.
