AI-Generated Fake Images Used for Refund Fraud on E-Commerce Platforms in China
Reportedly, users in China utilized AI image generation tools to create fake photographs of damaged e-commerce products, submitting them as evidence to obtain fraudulent refunds from sellers. The AI-generated images were reportedly difficult to distinguish from authentic damage photographs, posing challenges for merchants and platforms to detect and prevent the fraud.
What happened
Users on Chinese e-commerce platforms reportedly used AI image generation tools, particularly ChatGPT, to create fake images of damaged products for refund fraud schemes. According to reports, individuals purchased items such as durian fruits and photographed them in normal condition, then used AI tools to digitally alter the images to show damage such as mold or deterioration before submitting refund requests marked 'refund only' on e-commerce platforms.
Fraudsters also reportedly created AI-generated fake images of damaged shoes, clothing with tears, and swollen power banks. Online commentators and bloggers noted that the AI-generated images had become difficult to distinguish from authentic damage photographs, though some examples contained visible AI artifacts.
Reported impact
E-commerce merchants reported significant challenges in managing refund requests, as they could no longer easily identify fraudulent claims using traditional photo manipulation detection methods. The scheme reportedly enabled individuals to obtain refunds or free products without legitimate cause, directly harming sellers' economic interests.
Index analysis
The incident demonstrates the misuse of generative AI tools to circumvent platform fraud detection systems. The scalability of AI-generated content creation enabled coordinated or copycat fraud schemes that exploited the difficulty in distinguishing synthetic images from authentic product photographs.
Governance controls
E-commerce platforms were noted to face challenges in developing adequate technical and procedural safeguards against AI-generated fraud evidence, as traditional image forensics became insufficient.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.