TikTok ads use AI clones of popular figures Dewy Choo, Zhang Linghe to target S’poreans

SINGAPORE – Five years of filming and posting videos online have exposed Dewy Choo to much artificial intelligence-generated content on the internet – yet it still felt surreal when the social media influencer came face to face with one such video of herself on TikTok in July.

Without her consent, an anonymous user had hijacked her likeness and her voice to create an advertisement selling false eyelashes for an online store on the social media platform.

“I kept watching the video over and over, trying to examine every detail of her face – and I realised her eyebrows were exactly like mine,” Choo told The Straits Times.

“Her jaw, cheeks, and down to how her lips were dehydrated like mine – I was like, oh my God, that’s me.”

Choo has nearly half a million followers across TikTok and Instagram, where she posts frequently. The 24-year-old is among popular figures whose identities have been exploited to market products to Singaporean TikTok users, joining the likes of Chinese actor Zhang Linghe.

Ads featuring AI-generated images of Zhang originated from the same anonymous account. These advertisements depicted the actor at local landmarks, such as East Coast Park and Marina Bay, to sell products including mascara and caps. Two such videos seen by ST went so far as to replicate fine details, including a distinctive mole on Zhang’s nose.

Choo, who reported the AI-generated ad of her to TikTok, said that she had promoted the same brand of false eyelashes. So such AI dupes could hurt her work by diverting attention from her own commissioned videos. The dubious video was taken down following reports from her and her followers.

She was also worried that abusers may use her likeness to make malicious content. “When I showed the video to my boyfriend, he said the voice sounded exactly like mine,” said Choo.

Such content can be considered inauthentic material abuse under the Online Safety (Relief and Accountability) Act, which was passed in November 2025.

The newly minted Online Safety Commission (OSC) is empowered by the Online Safety (Relief and Accountability) Act to help victims of online harm through content takedowns and account restrictions.

The OSC, which became operational in June, is currently focused on tackling intimate image abuse, image-based child abuse, doxxing, online harassment and online stalking. It plans to address inauthentic material abuse in the months to come.

However, victims can still make a report to the agency if they believe that the deepfake material may fall under one of these five harm categories, said an OSC spokesperson in response to ST’s queries.

“When the provision (for inauthentic material abuse) comes into force, victims may report deepfake material to the OSC if it has caused them harassment, alarm, distress, or humiliation because of the false or misleading depiction,” said the spokesperson.

“Each case will be assessed based on its own facts and circumstances.”

When the provision for inauthentic material abuse comes into force, victims of such harms will also be able to sue perpetrators in civil courts for monetary damages which the court finds equitable, said Gilbert Leong, senior partner and head of Dentons Rodyk’s intellectual property and technology department.

However, the material has to be realistic enough such that a regular person would believe that the victim was actually involved in it.

“AI-generated material that is obviously fake or digitally generated may not fall under inauthentic material abuse,” said Leong.

If a celebrity’s likeness is used to sell products without their consent, they may also be able to sue under the tort of passing off, which protects the goodwill associated with a well-known person, said Gordon Lim, litigation partner at Focus Law Asia.

Saving copies of the offending material and related information such as the URL and date posted is helpful in making reports to the OSC and for potential legal claims, said Leong.

He added that those with brand or talent agreements – such as celebrities – should ensure that these documents expressly address derivative use of their likeness, image, or voice.

Where possible, users should also limit access to their social media accounts to people they know, said Lim.

AI-generated videos of popular figures make up just a fraction of a deluge of AI-generated advertisements that have flooded social media platforms and seek to target the local user base with characters that attempt a Singaporean accent and use Singlish terms.

ST found several anonymous accounts – with each having posted over 100 AI-generated ads – within an hour of scrolling on TikTok. The quality varies greatly, with many producing gibberish speech and text and inaccurate portrayals of the Singapore landscape.

“The use of AI tools and sloppy output alone does not necessarily cross the threshold of unethical advertising,” said Bryan Tan, chairman of the Advertising Standards Authority of Singapore.

Examples of unethical advertisements include claims that are not accurate representations of the actual product or location, as well as demonstrations, reviews, and testimonials from real and fictitious people who have never been customers, said Tan.

He added: “Advertisers should disclose the use of AI clearly if the absence of such disclosure may result in consumers being misled. However, the presence of a disclosure is not enough to correct false claims or representations in an advertisement.”