Published on November 2, 2023, 6:46 am
In the ever-evolving landscape of technological advancements, China has taken a significant step forward by actively seeking public opinions on national security requirements for generative artificial intelligence (AI) services. With the global shift from analytical AI to generative AI, Chinese AI firms are increasing their efforts to develop large language models like ChatGPT. The response to this technological shift varies across China, from excitement and acceptance to concerns about potential job displacement.
To set new standards and ensure the security of generative AI services, China’s National Information Security Standardization Technical Committee recently issued the Basic Requirements for Security of Generative Artificial Intelligence Services. This document provides guidance on addressing security issues related to GAI services, covering aspects such as training data security, model security, security measures, security evaluation, filing applications, and security assessments.
The Draft Requirements outline the essential security needs for GAI services and apply to both organizations and individuals providing these services within China. These requirements allow for self-assessment by GAI service providers or assessments conducted by third parties. They also serve as a reference for relevant regulatory authorities to evaluate the security of GAI services.
One crucial aspect addressed in the Draft Requirements is training data security. GAI service providers are required to avoid using illegal or harmful information during the training of AI models. Moreover, they must respect the legitimate rights and interests of third parties, including data rights, intellectual property rights, and personal information rights.
Another significant focus area outlined in the Draft Requirements is keyword libraries in GAI services. These libraries should include an extensive list of keywords that cover various security risks. The libraries are expected to house at least 10,000 keywords that don’t exceed 10 Chinese characters or 5 words in any other language.
Data rights protection is also emphasized in the Draft Requirements. GAI service providers must refrain from using data with conflicting rights or unclear origins and should possess proof of the legality of their data sources.
Model security is another important aspect covered in the Draft Requirements. AI service providers are required to establish content security, transparency, accuracy, and reliability. Additionally, the Draft Requirements outline seven essential security measures that should be followed by AI service providers to promote safety, transparency, and regulatory compliance.
To ensure responsible and safe deployment of generative AI services, security assessments play a crucial role. Providers are expected to conduct comprehensive security assessments that cover corpus safety, generated content safety, and question rejection. The results of these assessments should be supported by evidence and take into account all the provisions laid out in the Draft Requirements.
In conclusion, the Draft Requirements for Security of Generative Artificial Intelligence Services aim to enhance the level of security in GAI services while providing guidance to service providers and regulatory authorities. By establishing clear security standards and procedures, the Chinese government intends to ensure the safe and responsible deployment of generative AI services.
For more information on this topic: [China Seeks Public Opinion on Security Requirements for Generative AI Services](https://bnn.network/tech/ai-ml/china-seeks-public-opinion-on-security-requirements-for-generative-ai-services/) by [BNN Breaking](https://bnn.network).