With the rapid development of artificial intelligence technology, large models are increasingly used in various fields, but the security issues that come with it are also becoming increasingly prominent.
In order to build a solid artificial intelligence security defense line and improve the security risk prevention capabilities of large model technology, the China Academy of Information and Communications Technology (hereinafter referred to as "Information Academy") and 360 launched the development of the large model security benchmark SafetyAI Bench.
The large model security benchmark test aims to establish an authoritative Chinese benchmark test system for large model security in the industry. It adheres to the three principles of "fairness and justice, industrial application, and selection reference" and uses scientific and rigorous testing methods to evaluate the content security and data security of large models. Performance in safety and technology ethics.
The system will cover Chinese evaluation data sets in 27 dimensions, including correct values, legal compliance, privacy protection, civility and health, and AI autonomy, to provide the industry with comprehensive and detailed security performance testing indicators.
As an important research institution in China's information and communications industry, the Academy of Information and Communications Technology has profound research foundation and rich industry experience in the fields of artificial intelligence and big data. It is the only domestic ICT field that can connect new technology research, standard development, test verification, testing and certification to The industry promotes an industrial innovation platform for the entire chain and all links.
Its professional research team, extensive industry cooperation network and profound industry influence provide a solid guarantee for the smooth advancement of large model security benchmark testing.
360 Group has been deeply involved in the field of network security for many years, and has the world's leading security big data scale; and as an early domestic enterprise to deploy large-scale artificial intelligence models, 360's self-developed cognitive general-purpose large model - "360 Intelligent Brain" in Chinese It ranks first in domestic large model capabilities in multiple third-party evaluations such as the General Large Model Benchmark Evaluation.
With its deep accumulation and leading strength in digital security and artificial intelligence, 360 will provide strong technical support, practical experience and innovative ideas for this standard testing work, and jointly promote the establishment and improvement of the large-model security benchmark testing system.
Industry insiders said that the joint large-model security benchmark test between the Academy of Information and Communications Technology and 360 is a strong alliance with complementary advantages. It will not only establish an authoritative and fair security assessment standard in the industry, but will also further promote the innovation and development of large-model technology. application and contribute to the steady development of the artificial intelligence industry.
It is understood that Zhou Hongyi, founder of 360 Group, will also hold the first free AI class in the first year of 2024 large model application scenarios on February 29. At that time, the development trend of large model technology, as well as its application prospects and challenges in various fields will be discussed in depth.