PJ warns of fake celebrity voices and images created by AI

Recent deepfake videos featuring prominent figures – crafted from publicly available interviews to promote fraudulent investment schemes – have emerged alongside counterfeit websites mimicking local newspapers, leading to two reported AI-related fraud cases since April.

Among the cases, the Times was previously impersonated by scammers who created fake content featuring former Chief Executive Ho Iat Seng in a deceptive “interview.”

This fraudulent scheme lured the public into making investments, resulting in significant financial losses. In June this year, a similar scam reappeared on social media platforms in Macau.

Also previously, the scammers used news clips of the current Chief Executive Sam Hou Fai attending a public event, combined with AI-generated deepfake technology to fabricate false videos.

These videos impersonated the chief executive, promoting fraudulent online investment projects aimed at deceiving the public into logging into bogus investment platforms.

“AI has become an integral part of daily life, providing numerous conveniences to the public. However, criminal gangs are increasingly exploiting AI technology to commit or facilitate crimes. This has led to a rise in the sophistication and complexity of fraud schemes, significantly enhancing their success rates and posing a greater threat to the public,” the Public Security Forces Affairs Bureau said, acknowledging that AI-driven contactless crime is expected to become one of the mainstream trends in the future.

According to the bureau, fraudsters can leverage deepfake technology to impersonate anyone, creating fake videos of relatives or friends who urgently claim to be seriously ill or involved in an accident to solicit emergency funds.

They may also impersonate superiors requesting commercial transfers from subordinates. In these scenarios, not only ordinary citizens but also multinational companies and small and medium-sized enterprises (SMEs) can fall victim to AI-driven fraud.

Authorities further noted that AI technology permeates every stage of criminal activity.

During the preparation phase, AI can automate the collection and analysis of online information, identify potential victims, and generate the voice and text content necessary for scams. In the execution phase, AI chatbots can simulate “customer service representatives” or “romantic partners” to maintain regular contact with victims, gradually building trust.

Additionally, in terms of technical support, AI can automatically identify system vulnerabilities to facilitate cyberattacks or quickly create phishing websites.

In response to the rapid development of AI technology, distinguishing between truth and fiction in the online world has become increasingly challenging, particularly regarding photos and videos.

The Judiciary Police (PJ), responding to recent inquiries from the Times, emphasized that while professional software and tools can assist in verification, the public can also employ simple methods for preliminary assessments.

These methods include checking for blurry edges around subjects, observing if facial expressions appear stiff or unnatural, and noting any unusual distortions in the background.

Additionally, individuals should assess whether the direction of light or shadows seems unreasonable and listen for unnatural voice quality or mismatched lip movements. If there are doubts about the authenticity of a photo or video, it is advisable to verify the information from multiple sources.

The police also reminded the public that technological advancements are rapid, and any auxiliary identification techniques or tools may quickly become outdated. Therefore, individuals should not blindly trust the content of images or videos circulating online.

According to the PJ, they have implemented several measures to combat emerging threats. These include enhancing criminal investigation capabilities, strengthening intelligence gathering and online monitoring, and improving interregional and international cooperation to quickly inform the public about new criminal tactics.

Additionally, they are enhancing communication with major social media platforms to swiftly address and remove reported illegal or fraudulent content, thereby protecting the public’s rights and interests.

The “Macau Residents’ Internet Usage Trends Report 2025” reveals that nearly 80% of internet users are aware of “generative AI,” with about 50% having used it.

DeepSeek is the most commonly used tool, followed by ChatGPT. Users primarily utilize generative AI for information retrieval and problem-solving, with an average effectiveness rating of 3.7 out of 5 for learning and work.

Additionally, nearly 60% of users expressed a willingness to use generative AI in the future, indicating a growing intent to adopt this technology.

The PJ reminded the public that using artificial intelligence technology to disseminate false information that harms others’ reputations or to defraud individuals may constitute a crime under existing laws.