A Taiwanese male model accuses Life8 clothing of “stealing faces with AI”! The brand responds with automatically generated content

A Taiwanese male model posted an accusation against the clothing brand Life8, saying it generated AI product images using his facial appearance features without his consent and put them on its official website to sell clothes. The AI duplicated his features before surgery—his single eyelid, the right side of his neck tattoo, and his usual standing pose. Life8 first responded with: “The AI was generated according to instructions, and it is not created for a specific individual,” and then it privately messaged him to discuss authorization of portrait rights—contradicting itself and sparking a second wave of backlash.
(Background summary: YouTube launched an AI deepfake detection tool: creators aged 18 and above can detect portraits and submit takedown requests)
(Additional background: Sony is bringing Suno to court this month: whether using AI to train on copyrighted songs counts as “fair use” is being argued in court for the first time)

Table of contents

Toggle

• Single eyelid and tattoo—caught in the act
• Chinese e-commerce: a few tens of cents per image
• If you really sue, the key is not copyright

Key takeaways

• A male model accuses Life8 of stealing his face to generate AI product images; the single eyelid and neck tattoo are key tells
• In Chinese e-commerce, the cost of generating an AI image is so low that it’s just a few tens of cents per image; up to 8,000 images can be produced in an hour
• Taiwan’s main battleground is Article 18 of the Civil Code—portrait rights; photo copyright is not something the model owns in the first place

Recognizing himself was easy—the single eyelid he had before surgery. On the 17th, a Taiwanese male model surnamed Lin (social account iamgalin) posted a video, accusing the clothing brand Life8 of generating AI portrait images using his appearance features without his consent, and putting them directly on its official website to sell products.

View this post on Instagram

The tells listed by the model surnamed Lin were very specific: the AI image kept his single eyelid before surgery, and the neck tattoo location on the right side matched exactly, down to the same stance and the same head angle—also the habitual moves he makes during his work. “I was replaced by AI using my own photos.”

What’s more troublesome is the consequence. He said that if customers see the website, they might think he had undergone plastic surgery, and therefore decide not to work with him. He demanded Life8 take down all images and provide a reasonable explanation.

Single eyelid and tattoo—caught in the act

Life8 initially responded in social media comments, saying the images were generated by AI according to instructions and were not created with any specific person as the subject. It said it did not deliberately use any specific person’s photos or data, and stated it had stopped using the relevant images, re-examined the content, and apologized for the trouble.

This statement didn’t put out the fire—it only made it burn hotter. Later, the brand also told the model in comments on his community posts that it had privately messaged him to handle portrait rights and subsequent authorization matters. Netizens immediately seized on the contradiction: it had just said it wasn’t using any specific person as the subject, so whose authorization was it discussing afterward?

Pushing the responsibility onto “AI generated according to instructions” is the weakest argument, because the instructions are given by people, and the reference images are also chosen by people. The failure point in this PR battle isn’t the technology—it’s that the two sentences before and after don’t match.

After the incident escalated, other models also shared similar experiences, showing this isn’t a one-off case involving a single brand. If the parties were truly to file a lawsuit, this could very possibly become Taiwan’s first AI portrait infringement lawsuit.

Chinese e-commerce: a few tens of cents per image

Turning the camera to the other side of the strait, you can see that the issue being hotly debated in Taiwan is something Chinese clothing e-commerce already went through.

In China, the cost structure is such that, for instance, a typical model is paid per item at about 100 to 300 yuan per piece. Add to that the photographer, makeup artist, and venue fees—shooting dozens of styles in a day makes it easy for expenses to exceed ten thousand yuan. Swapping to AI tools can reduce the cost of a single image to only a few tens of cents, with image generation taking only a few minutes.

The savings aren’t only the model’s money, but also the money for the entire supply chain—photographers, makeup artists, lighting, and venues. For example, the following well-known AI model workflow has been spreading in China for more than half a year already.

In general, China’s apparel industry says the efficiency gap is enormous: with AI, there’s no need to secure booking dates, wait for models, or wait for the weather. Industry descriptions say that in an hour, one can produce 8,000 retouched images. Even if 99.99% are junk, picking out a single usable one is still profit. For fast-fashion merchants with very high new-release frequency, styles that used to take half a month to shoot can now be handled in a single day.

On the platform side—such as Alibaba—on January 15 of this year, it announced connecting Qwen to Taobao, opening AI shopping service beta testing to all users. The Taobao app has also fully deployed an AI try-on service, enabling users to try on and mix-and-match multiple clothing items.

However, the technology wasn’t quite ready. After Taobao’s AI try-on went live, users complained because they expected to see “themselves” wearing that outfit, but what they actually got was the same catalog model with just a slightly adjusted jawline. The speed at which real models get replaced is far faster than the speed at which consumers’ experience is met.

If you really sue, the key isn’t copyright

Back in Taiwan, the most common misunderstanding in cases like this is thinking the model is asserting photo copyright. In reality, the model doesn’t own the photo copyright.

Under Taiwan’s Copyright Act, the copyright in a photographic work generally belongs to the actual creator—the photographer. Article 11 provides that an employed photographer is the author, and the employer is the holder of the copyright property rights; Article 12 provides that an independent contractor photographer is the author and the holder of the copyright property rights, and that the funding party has the right to use the work. The person being photographed is not the creator, and thus not the author.

What the model can truly claim is personality rights. Taiwan’s Civil Code doesn’t explicitly write the term “portrait rights,” but it treats it as one of the personality rights protected under Article 18(1) of the Civil Code. It then pairs that with Article 195 to request an injunction to stop the infringement, restoration of reputation, and damages.

The core test is the scope of use. Agreeing to being photographed once does not mean agreeing to use the portrait for other purposes. Once the use exceeds the scope of utilization that the photographed person could reasonably have expected at the time of being photographed, there is room to assert rights.

So does copyright law have no role at all? Not exactly. What you can file a complaint against is the brand that previously commissioned and shot the photos. If someone intentionally feeds photos protected by copyright into AI for training, that is a “reproduction” under copyright law, and in principle authorization is needed, unless it falls under fair use.But that right belongs to the copyright holder of the photos—the photographer or the brand that funded the shoot—not to the model himself/herself.

In other words, if a business uses another brand’s photos to train or as references, and you can sue over copyright, the target could be other brands rather than the person who appears in the photos.

There’s also an angle that receives less attention. Original tattoo designs in Taiwan are considered fine art works, and the copyright in principle belongs to the tattoo artist. If the AI image fully copies that design, it could theoretically also infringe on the tattoo artist’s copyright—separate from the model’s portrait rights, as two independent issues.

Ironically, in its online response, Life8 cited that according to an MOEA Intellectual Property Office interpretation, whether an AI-generated output is protected by copyright depends on whether there was actual human creative input in the creation process. If the output is completed entirely by independent AI computation and the user only gives simple instructions without substantial creative input, the output is not protected by copyright.

If Life8’s claim—“AI generated according to instructions, not created with any specific person as the subject”—holds true, then it also effectively concedes that these product images may not have any copyright at all, making it difficult to assert rights even if the entire set gets copied and used elsewhere by competitors in the future.

Sounds like it’s hard to regulate the legal source of AI training? Taiwan’s regulations are only now catching up on that front. On December 23, 2025, the Legislative Yuan passed the Artificial Intelligence Basic Law in third reading, comprising 20 provisions. The central competent authority is the National Science and Technology Council. The law stipulates seven core principles that government promotion of AI should follow: sustainable development and well-being, human autonomy, privacy protection and data governance, cybersecurity and safety, transparency and explainability, fairness and non-discrimination, and accountability. It also requires appropriate information disclosures or labeling for AI outputs.

But this is a “basic law.” Its nature is top-level guiding principles; the specific penalties and enforcement details revert to the individual acts administered by each competent authority. That means that in the short term, these disputes still have to be handled under the Civil Code personality rights, the Copyright Act, and the Personal Data Protection Act—there won’t be automatic new penalties just because the basic law exists.

Common questions

If a model finds their photos are used to generate AI images, what can they sue for?

The main battleground is portrait rights. The photo’s copyright under Articles 11 and 12 of the Copyright Act usually belongs to the photographer or the funding party; the model is not the author. The model may request an injunction to stop the infringement and damages under Civil Code Article 18 and Article 195. The key is whether the use exceeds the scope agreed at the time.

Does Taiwan require AI-generated images to be labeled?

On December 23, 2025, the Legislative Yuan passed the Artificial Intelligence Basic Law in third reading, comprising 20 provisions, which requires appropriate information disclosure or labeling for AI outputs. But because it’s a basic law, the specific penalties and details are set separately by each competent authority under its own laws.