China's AI Governance Plan Cuts Deepfake Attacks by 60%
China's AI Defense System Gains Global Recognition at UN Summit
At the recent Global AI for Good Summit in Geneva, China demonstrated significant progress in combating deepfake threats, particularly in financial sectors. Peng Jin, Vice President of Ant Group's Technology Strategy Department, revealed their system has reduced facial recognition attacks by 60% while maintaining 99.9% accuracy.
Peng Jin presents China's anti-deepfake achievements at the UN summit
The Global AI Security Challenge
The summit, co-hosted by the International Telecommunication Union (ITU) and Swiss government, brought together UN agencies to address AI applications for sustainable development. As generative AI tools proliferate, financial institutions face growing threats from sophisticated identity forgery techniques.
"In 2024, we saw deepfake attack rates peak at 10% in Southeast Asian banks," Peng explained. "Through our multi-layered defense system, we've reduced this to 4% while slashing false rejection rates from 50% to 3%."
How China's System Works
Ant Digital's subsidiary ZOLOZ combats two primary attack vectors:
- Presentation Attacks: Using static images or videos to bypass liveness detection
- Injection Attacks: System-level camera hijacking to insert pre-recorded footage
The solution combines:
- Device-level environmental safety checks
- Dynamic deep learning models for biometric verification
- Comprehensive risk control strategies
Currently deployed across 25+ countries, the technology represents one of the most robust financial-grade authentication systems available.
Advancing Global AI Security Standards
In a significant move for transparency, Ant Digital recently released:
- 1.8 million multimodal deepfake datasets
- Coverage of 88 forgery techniques
- Shared at IJCAI 2025 to foster verifiable AI research
The daily algorithm updates demonstrate an ongoing arms race against increasingly sophisticated forgery methods emerging on black markets.
Key Points:
- 60% reduction in deepfake attacks achieved through Chinese technology
- System maintains 99.9% accuracy with minimal false rejections
- Solution now protects financial institutions across 25+ countries
- China sharing datasets to advance global AI security research
