China Tests AI Models for Compliance with Core Socialist Values
- Chinese government officials are rigorously testing AI models to ensure they align with core socialist values, expanding the country's censorship regime. - The Cyberspace Administration of China (CAC) is conducting mandatory reviews of AI models from major tech companies like ByteDance, Alibaba, and start-ups such as Moonshot and 01.AI. - The review process includes batch-testing AI responses to politically sensitive questions and examining training data and safety protocols. - China's stringent approval process has led AI developers to implement extensive censorship measures, including filtering training data and building sensitive keyword databases. - Despite the strict regulations, Chinese officials aim to avoid AI models that completely dodge political topics, setting limits on the number of questions AI can decline during safety tests. Source: Financial Times
jinyang
Stealth
2 months ago
Discover More
Curated from across
