China works on AI safety benchmark as regulators target large model risks
• China has launched a state-led initiative to develop an AI safety benchmark aimed at mitigating risks associated with large language models (LLMs). • The system will utilize automated fuzzing, stress testing, and human oversight to reduce data leaks, control hallucination rates, and prevent "jailbreak" attacks. • This move signals Beijing's commitment to joining global efforts to enhance the oversight and security of generative AI technologies.
scmp.com


