China has issued a finalized national standard, GB/T 45654—2025, titled "Cybersecurity Technology – Basic Safety Requirements for Generative Artificial Intelligence Services," on May 28, 2026. This standard aims to enhance the safety and security of generative AI (GenAI) services, particularly those with public opinion-shaping or social mobilization capabilities.
It specifies requirements for training data safety, model safety, and general safety measures. The final version includes detailed best practices, such as establishing procedures to prevent the inclusion of personally identifiable information or copyrighted works in model training data without explicit permission. This addresses new cybersecurity risks arising from widespread GenAI applications. The standard applies to service providers and guides oversight departments and third-party evaluators, though its enforceability faces challenges from rapid innovation and competition within China's AI industry.
No comments:
Post a Comment