news

jimu news reporter zhou dan

on september 6, at the 2024inclusion·bund conference "protecting ai with ai, the offense and defense in the big model era" forum, wdta world digital technology academy officially released the international standard "big model supply chain security requirements". the standard was jointly compiled by experts from dozens of domestic and foreign units including cloud security alliance (csa) greater china, ant group, microsoft, google, meta, privateai, etc. this is the industry's first international standard for big model supply chain security, marking another solid step forward in international cooperation in global ai governance.

peter major, executive chairman of the united nations commission on science and technology for development and founding chairman of the world digital technology academy, who attended the meeting, said that as artificial intelligence continues to develop and affect various industries around the world, international cooperation on artificial intelligence-related standards has become increasingly important. the development of globally recognized international standards for artificial intelligence will help cultivate trust in artificial intelligence, promote innovation, and ensure that artificial intelligence systems operate reliably in different industries and regions.

regarding the standards released this time, li yuhang, chairman of the cloud security alliance (csa) greater china, said that the operation of large model systems depends on the interconnected supply chain ecosystem distributed around the world. the widespread application of global artificial intelligence large models has brought unprecedented opportunities, but also exposed huge challenges in supply chain security. the release of this standard by wdta provides a systematic management framework for the full life cycle security of large models, solves security challenges in the supply chain, supports the security and sustainable development of global artificial intelligence technology, and enhances the overall credibility of the industry.

information shows that "large model supply chain security requirements" is one of the wdta artificial intelligence security, trustworthiness, and responsibility (ai str) series of standards. this standard provides a supply chain security protection framework for large language models, and provides requirements for supply chain-related security risks and supply activity management involved in all aspects from data preparation, large model development to deployment and operation, and provides common supply chain security risks, typical security cases and other related information. through this standard, potential supply chain security risks faced by large model systems in their life cycle can be effectively identified and evaluated, such as data leakage, model tampering, and supplier non-compliance, to ensure the integrity, availability, and confidentiality of the supply chain, thereby improving the security of large model systems.

it is understood that this standard not only provides a systematic security management framework for both supply and demand sides in the large model supply chain, but also provides a reliable basis for security review and compliance management of third-party organizations and authoritative departments, further enhancing the overall security and reliable development of large model systems.

the world digital technology academy (wdta) was established in april 2023 and is an international non-governmental organization registered in geneva. the organization is committed to advancing digital technology and promoting international cooperation on a global scale. the ai ​​str program is a core initiative of wdta, which aims to ensure the security, trustworthiness and responsibility of artificial intelligence systems. microsoft, google, anthropic, ant group, baidu and others are all its members. at the united nations science and technology conference in april this year, wdta released two large-scale model security international standards compiled by dozens of units including openai, ant group, iflytek, google, etc., both of which are ai str series standards.

in addition to the release of the standard, the forum also announced the official launch of the "global ai attack and defense challenge", the first domestic technology competition themed on large-scale model attack and defense. the competition focuses on the industrial practice of ai large models, and has designed two-way tracks of "attack" and "defense". it invites technical talents from all walks of life to compete in the security of large models themselves, anti-counterfeiting detection of large model generated content, and large model abuse risk detection, in order to compete for a million-yuan technology prize.

it is understood that this event is co-organized by the chinese society of image and graphics, ant group, and cloud security alliance (csa) greater china, and jointly initiated by c9 universities such as tsinghua university, shanghai jiaotong university, zhejiang university, and many industry-university-research organizations. as a technical cooperation unit, the shanghai artificial intelligence laboratory aims to gather the strength of academia and industry, face and solve the potential risks in large model applications, and help the global ai industry develop healthily and sustainably.

(source: jimu news)

for more exciting information, please download the "jimu news" client from the application market. please do not reprint without authorization. you are welcome to provide news clues, and you will be paid once adopted.