AI-driven refund scams spur calls for stronger e-platform protection

Small online businesses are facing a troubling trend — customers submitting artificial intelligence-altered images of damaged goods to claim refunds, prompting a law expert to call for stronger platform protections against digital fraud.

For three months, an online fresh produce store in Jiaxing city, Zhejiang province, has been a victim of several refund-only scams, potentially orchestrated using AI technology, China Central Television reported in November.

In October, Chen Xiaowei, who works in Shanghai, returned to her hometown to investigate the issue for her father, the store owner.

"My father's store, known for picking and shipping fresh citrus fruits, has been in operation for five years. He records videos of each package before delivery, documenting the color, size, and packaging serial number of the fruits," Chen told CCTV.

Despite these precautions, Chen discovered that the buyers involved in the refund claims had submitted unboxing photos showing the fruits as either rotten, shriveled, or infested with bugs.

Her father initially attributed the problems to delivery and agreed to refund the first few claims. However, as the incidents multiplied, his efforts to resolve the issue with the courier company proved futile, prompting him to seek his daughter's assistance.

"I noticed a watermark-like trace in the corner of one photo, suggesting it had been altered," Chen was quoted as saying.

Chen submitted the packaging videos and suspected AI-modified images to the online platform, but her complaint was rejected due to insufficient evidence. The platform suggested she negotiate directly with the buyers, but the individuals involved typically blocked further contact.

Efforts to seek assistance from the platform's service center were met with the response that they could not determine if the images were AI-generated. With the total order amount falling below 1,000 yuan ($142), it did not meet the threshold for filing a fraud case.

Chen expressed frustration at the ease with which AI-generated watermarks can be removed and criticized the platform's lack of detection mechanisms, which leaves honest small business owners like her father vulnerable to related scams.

Similar scams

In a similar incident reported by Legal Daily in November, an online toy store owner from Guangxi Zhuang autonomous region received a buyer's request for a refund without returning the goods for a supposedly damaged 50-yuan plush toy.

The buyer sent a photo showing the toy looking worn and dirty. Suspicious, the owner used an AI detection tool, which revealed signs of AI manipulation in the image.

When the owner challenged the buyer, he was threatened with consumer complaints. The platform initially sided with the buyer. The owner's appeal was only successful after he shared his story on social media, attracting widespread attention.

Legal Daily revealed that similar AI-manipulated refund claims are common. Items, especially like food and toys, are often altered to show damage, with some images even showing AI-generated watermarks.

Zhao Jingwu, an associate professor of law at Beihang University, said that AI-generated refund fraud primarily targets fresh produce and food items because AI can create deceptive images and videos of these products that are hard to detect. Although rare now, this type of fraud could expand to other online products with advancements in AI technology.

Legal view

According to the Civil Code, using AI to falsify evidence for refunds is considered fraud. This law allows businesses to seek compensation for economic losses. If the amount involved doesn't meet the criminal threshold, administrative penalties like fines or detention may apply. For larger sums, it could be classified as criminal fraud under the Penal Code, Zhao said.

In a CCTV report, an online store owner surnamed Gao from Suzhou, Jiangsu province, sent a box of crabs in November to a customer, who shortly after delivery claimed that six out of eight crabs were dead, providing suspicious evidence to back up the claim.

After compensating the buyer 195 yuan, Gao noticed discrepancies in the number of male and female crabs between the photos and videos the customer provided and shared her doubts online, prompting a police investigation. It revealed that the buyer had used AI to fake the video, resulting in an administrative detention of eight days and recovery of the fraudulent refund.

"Platforms are not required to thoroughly verify user-submitted evidence, as they lack knowledge about the transaction. However, they should alert merchants to potential AI fraud risks if images in refund requests show clear signs of being AI-generated," Zhao said.

"Merchants can deter AI-driven refund fraud by requiring multiple specific actions in the video and longer video submissions from customers. Current AI technology struggles with seamlessly integrating human features and maintaining video fluidity, which can be exploited to detect fraud," he said.

Zhao added that since the measures for identifying AI-generated synthetic content took effect on Sept 1, most domestic AI products include explicit and implicit markers in generated content. However, fraudsters may use overseas AI tools or remove these markers. By using technical methods to remove explicit markers, it becomes challenging for platforms and merchants to verify the authenticity of the information through manual review.

"To combat this, platforms should include AI-generated synthetic content detection mechanisms in refund processes and support merchants with comprehensive video documentation of packaging and delivery. Platforms need to regularly update merchants on new AI fraud detection methods, such as identifying anomalies in lighting or body positioning," Zhao said.