중국 동영상 ‘철통심사’의 역설…AI 허점 파고든 ‘계정 때리기’ 산업

A hidden paradox emerges from China's highly regulated short video platforms, as artificial intelligence (AI)-driven content moderation systems are being exploited to create an "account crushing" industry. In recent reports from Chinese media, it's revealed that while AI-powered content auto-screening efficiently removes rule-violating material, the verification process often falls short. This loophole has allowed a thriving underground market where video deletions and account bans can be bought and paid for.

Companies and brands use this "打号·计划时代" (da huo, or "hammering accounts") service to remove content that might harm their reputation. Reporters claim that one influencer with 260,000 followers had their account banned, causing other accounts to be reported as well. The reporter's account was threatened with more bans if they didn't submit additional reports, and after sending 1000-point evidence, their own video was restored.

The 'account crushing' trade has already established itself as a market force, with online advertisements offering promotions like "25-point accounts (around 5,000 yuan) for 500 reports" or "1200-point accounts (around 24,000 yuan) guaranteeing video deletion." One account crusher claimed to have earned 53,715 points (around 110 million yuan) in a single month. These services often involve ordering from overseas social media accounts and disseminating the system and language across different platforms.

AI screening systems' vulnerabilities are the key to profiting from this scheme. Sellers fabricate stories of privacy breaches or copyright violations, submit forged AI-generated data, and submit fabricated reports. When the Beijing News interviewed a reporter about a video they submitted for 'privacy violation' that was later removed, the reporter claimed the video was removed within two hours without proper verification.

Meanwhile, the victim had to provide identification and evidence and waited for days before their request was denied - but once a video is deleted, the original uploader can simply request its removal again.

The system is designed with both the ability to delete and recover videos, creating a powerful advantage for malicious reporters. A security expert notes that platforms often resort to automatically deleting AI-screened videos first and allowing a second look during compliance checks, creating a significant cost imbalance in favor of the aggressor.