Douyin VP Warns AI Fuels Rumor Spread, Unveils Countermeasures
Douyin Deploys AI Agents Against AI-Generated Misinformation
Beijing, October 29, 2025 - During CCTV's Social and Legal Channel special on AI-generated fake news, ByteDance Group Vice President Li Liang issued a stark warning: "While AI and algorithms are neutral tools, they've dramatically lowered the barrier for rumor creation." The executive identified rumor management as ByteDance's top priority for 2025.
Technological Counteroffensive
Li detailed Douyin's deployment of large language models to develop an "Intelligent Agent for Rumor Management" capable of:
- Scanning全网 (entire web) for authoritative sources within seconds
- Proactively identifying potential misinformation patterns
- Automating initial verification processes before human review
The system reportedly processes thousands of content pieces hourly, flagging suspicious claims about health emergencies, financial scams, and geopolitical events.
Limitations of Tech Solutions
"Technology alone cannot solve this societal challenge," Li cautioned during the broadcast. He emphasized three critical non-technological requirements:
- Media partnerships: Established news organizations providing real-time fact-checks
- Government collaboration: Official platforms sharing authoritative data streams
- User participation: Crowdsourcing rumor reports with contextual evidence
The VP cited recent cases where coordinated efforts debunked viral AI-generated videos about vaccine side effects and fabricated celebrity endorsements.
Industry-Wide Implications
This development comes as:
- Global platforms report 300% YoY increase in sophisticated AI-faked content (EU DisinfoLab Q3 2025)
- China's CAC implements stricter "Traceable Content Origin" regulations effective December 1st
- Competing platforms like Kuaishou invest in similar detection systems
ByteDance plans quarterly transparency reports detailing rumor interception rates and false positive ratios starting Q1 2026.
Key Points:
- 🔍 Douyin's AI agent cross-references claims against 10,000+ verified sources
- ⚖️ Current detection accuracy stands at 89% with <5% false positives (internal testing)
- 🤝 Partnership program with 45 state media outlets launched last month
- 📱 Users can now attach geolocation metadata to rumor reports for verification
