Künstliche Intelligenz wird zum Komplizen bei Online-Shopping-Rückzahlungsbetrugsversuchen

Artificial intelligence (AI) is used to create fake product defect images. Such "false" proofs are used for "asparagus hunting" and have even spawned a paid gray - area industry. Merchants are caught in a legal defense loop where "it is difficult to present evidence and even more difficult to file a complaint".

Since last year, "asparagus hunting" using AI - generated damage images in e - commerce has become a new gray - area industry.

Recently, Ms. Gao, the owner of an online shop selling Jiangsu hairy - crab milk, became a victim of a new fraud model. Of the eight crabs she sent, in the pictures and videos sent by the buyer the next day, "six were dead", and she was asked to refund 195 yuan. The police found that the video was AI - forged. The involved buyer was sentenced to eight days in prison. This is the first case in China where someone has received an administrative penalty for using AI - generated videos to obtain refunds in online shopping.

Ms. Gao's experience is not an exception. Reporters from "Science and Technology Innovation Board Daily" found in their investigation that with the spread of AI tools, the technical hurdles for forging images and videos of defective products have dropped drastically. From food to clothing, realistic forged "proofs" are used by a small group of people for "asparagus hunting" and have even spawned a paid training gray - area industry. Facing such fraud attempts, e - commerce merchants are caught in a legal defense loop where "it is difficult to present evidence and even more difficult to file a complaint".

AI - generated damage images for "asparagus hunting": The AI gray - area industry behind refund disputes

With the help of AI generation techniques, one can easily create forged refund proofs from a simple original image of an opened product, which are almost "indistinguishable from real ones".

Mr. Zhong, an e - commerce merchant with 15 years of experience in the porcelain industry, was also a victim of AI - generated image fraud last month. A buyer requested a refund only due to product damage and sent several pictures showing multiple cracks on the porcelain teapot. Despite his industry experience, Mr. Zhong was initially unsure if the pictures were real. But under the continuous pressure and accusations from the buyer, he wavered briefly. It was only when he firmly demanded that the buyer send a "damage video" that his colleagues in a professional group made it clear to him that it was an AI - generated forged video - "It looks fake when you look closely, it's AI". Thanks to his professional knowledge, Mr. Zhong recognized the fraud, and the buyer finally withdrew his request.

A year ago, the AI technology capable of generating realistic forged photos and videos was still the privilege of some technology companies and professionals. Today, with the spread of open - source AI tools and commercial AI services, the hurdles for this technology have dropped significantly.

Reporters from "Science and Technology Innovation Board Daily" noticed that since this year, merchants on various e - commerce platforms have reported on channels like Xiaohongshu that some consumers are using AI techniques to generate images or videos of defective products to file refund requests.

Such forged content includes various scenarios such as spoiled food, damaged clothing, and leaking cosmetics and is very realistic visually. In particular, food is a major target of such fraud attempts due to the difficult return inspection.

Lawyer Tu Panyue from Shanghai Guangming (Hefei) Law Firm also confirmed to the reporters that he had handled several similar cases six months ago. The fraud targets mostly focus on small items such as fruits, clothing, and shoes.

To confirm the phenomenon of AI - based forging, reporters from "Science and Technology Innovation Board Daily" also conducted their own tests using common AI tools like Tencent Yuanbao and Doubao. With simple instructions like "Show a part of the fruit in the picture as spoiled", the large - scale model could generate very deceptive "proof" images in 30 seconds.

AI - generated images of spoiled fruit

Although the generated images are usually watermarked with "AI - generated", this watermark is mostly in a corner and can be easily removed by cropping. Nevertheless, some merchants reported in their posts that many of the images presented as "proof" still had untreated watermarks directly on them, clearly showing the traces of AI - based forging.

According to the data released by "E - Complaint Treasure", the "refund problem" accounted for 16.16% and ranked first among the complaints of merchants on national e - commerce platforms during the "618" shopping event; the problem of "arbitrary refund claims" followed closely with a proportion of 12.10%. It is easy to see how widespread refund disputes are currently, and the abuse of AI forging techniques has undoubtedly increased the technical complexity in this gray area.

It is noteworthy that some merchants reported that even if they could recognize that the buyer had used AI to forge product damage, their complaints to the platform were often rejected due to "insufficient evidence", and they were finally asked to "negotiate with the buyer".

At the same time, since the individual or total amount of the related orders is often less than 1,000 yuan and thus does not reach the criminal threshold for fraud, it is difficult for merchants to turn to the police, and they are in a dead - end in legal defense.

With the rapid drop in the hurdles for using AI tools, the cost of using this technology for illegal or improper actions has also decreased. The low cost of abusing the technology and the high cost of legal defense have put many e - commerce merchants in a new difficult situation.

It is noteworthy that in addition to using AI to generate forged images, a gray - area industry around the "refund - only" approach has also emerged. Previous media investigations showed that for about 288 yuan in "course fees", one can learn the complete "art of the refund - only" approach from someone who calls themselves a "master", and it is claimed that a single account can achieve about 30 successful refunds.

When AI becomes a forging tool, how can we solve the liability problem behind small - scale fraud attempts?

The recent administrative penalty in China's first case of AI - generated videos in online - shopping fraud means that the law is now responding to this new form of crime. However, the current legal system is still lagging behind in the face of the rapidly advancing AI technology.

Lawyer Cai Bin from Shanghai Fuchen Law Firm told the reporters from "Science and Technology Innovation Board Daily" that he had received several inquiries from merchants regarding AI - based fraud attempts, but the amounts involved in these cases were usually small. Regarding legal defense, Cai Bin gave clear advice: On the one hand, one can turn to the platform; on the other hand, one can also turn to the police.

He pointed out that such actions meet the elements of fraud according to Article 266 of the Criminal Law. In practice, however, it is difficult to conduct a criminal prosecution because the amounts often do not reach the criminal threshold. He further explained that the police can sanction such actions in accordance with the Law on Public Order and Security, for example, with a prison sentence of five to ten days.

Technically, the "anti - AI" technology for preventing AI - generated videos is also developing. However, the spread and application of these technologies still require time and are costly. For most small and medium - sized e - commerce platforms and merchants, it is currently unrealistic to set up such technical defense systems on their own.

In an interview with the reporters from "Science and Technology Innovation Board Daily", several lawyers emphasized that combating the abuse of AI technologies in the e - commerce environment depends on the cooperation of multiple parties. Meanwhile, the e - commerce platform should play an even more important role and take on the core responsibility for improving technical inspections and optimizing rule - making.

"Currently, there is still a certain delay in establishing the 'refund - only' system and the risk - management system for detecting AI forgeries on large e - commerce platforms," said Lawyer Tu Panyue from Shanghai Guangming (Hefei) Law Firm directly. "For e - commerce platforms, handling the'refund - only' problem is not simply a 'judge' role. They have multiple inspection and due - diligence obligations."

At the same time, Tu Panyue suggested that the relevant regulatory authorities should encourage the platforms to organize a joint fight against black and gray industries such as "abuse of AI", "agent - related complaints and refunds", and "consultation and guidance for professional compensation claims".

Moreover, from a more macroscopic governance perspective, the establishment of new protective barriers should go beyond the platform level. Cai Bin further told the reporters from "Science and Technology Innovation Board Daily" that the legal and regulatory aspects also need to be improved to combat e - commerce fraud caused by the abuse of AI technologies.

He believes that the legislation needs to adapt to the times. Regarding new forms of crime committed with new technologies like AI, which are already on the rise, one should consider adding special articles to the Criminal Law or issuing judicial interpretations to create clear legal regulations.

At the same time, promoting consumer compliance education is essential. The technical tools themselves are not to blame; it depends on how they are used. Through basic education, consumers should be encouraged to clearly understand the boundary between legal and illegal use of the technology, so as not to abandon the technology out of fear of damage and to protect the innovation enthusiasm in the technology industry while combating crime.

In addition, Cai Bin suggested developing social conflict - resolution mechanisms. Conflict resolution should not be limited to courts, the police, or e - commerce platforms. In the consumer field, one could consider creating more diverse and earlier - stage out - of - court conflict - resolution channels to provide efficient solutions for conflicts that do not reach the criminal threshold.

This article is from the WeChat account "Science and Technology Innovation Board Daily", author: Li Jiayi & Tian Ye. Published with the permission of 36Kr.