This data security forum has made it clear, let alone underestimate its future destructive power. Don't overestimate the risks of ChatGPT | Data | Forum
On the second anniversary of the implementation of China's Data Security Law, the "Pujiang Escort" Data Security Shanghai Forum was held on September 1, 2023. The reporter learned from the forum that telecommunications and Internet enterprises carry a large number of user personal information and business data, accounting for more than a quarter of the total data generated by industry institutions nationwide, which is of high value, but also become the main target of network attacks. Security incidents such as extortion viruses and data leakage occur from time to time, and the overall situation is still grim. But among them, Shanghai plays an important role in safeguarding data security. According to Zhang Yuxiang, Deputy Director of the Shanghai Municipal Government Office, as of the end of last year, there were over 1200 digital core enterprises in Shanghai, with a core industry scale of nearly 340 billion yuan.
Since November last year, the release of new technologies and businesses such as generative artificial intelligence represented by ChatGPT has triggered industry changes, but it has also brought new challenges to data security protection and regulation. Ma Yunchao, Vice President of Shenxin Technology Co., Ltd., gave an example, "Our company has a detailed product rule library that requires four people to spend several years continuously updating and maintaining. However, the trained large model only took a week to surpass the original update and maintenance effect of the rule library. We are truly shocked."
But Cui Jiuqiang, the general manager of Shanghai Digital Certificate Certification Center Co., Ltd., is concerned about the "evil" side of the big model. "I have watched a scam short video where AIGC was used to generate a person's voice and portrait. The concept of 'seeing is believing, hearing is believing' is no longer possible. Facial recognition was one of the first commercially successful applications of artificial intelligence technology, but if AIGC starts to do bad things, such as various forgery attacks, the consequences will be huge." Zhang Zhaolong, founder of Guanan Information, also noticed that there are many sensitive data hidden in the big model. Not long ago, a technician from Samsung, South Korea, was struggling with coding and sought help from ChatGPT. They also uploaded the core code produced by the company, putting sensitive personal, corporate, and national data at constant risk. Therefore, it is crucial to do a good job in data leakage prevention and unified data control.
"From the heavyweight release of ChatGPT last year to the 'Hundred Model Battle' of domestic large-scale models since the beginning of this year, it can be said that it is lively and extraordinary. However, based on past experience, we should remember: never overestimate the creativity of a new technology in the first five years, and never underestimate its destructive power in the next five to ten years." said Wang Yanfeng, Vice Dean of the Institute of Artificial Intelligence at Shanghai Jiao Tong University.
The country and government departments have actually noticed the challenges involved. On August 31 of this year, the first batch of eight large model companies in China were put on record through the Interim Measures for the Management of Generative AI Services, and were approved to open their services to the public. These large models include the Scholar General Model of Shanghai AI Laboratory, Shangtang Technology "Discussion", and Baidu "ERNIE Bot". This means that in the process of transitioning from the "birth wave" to the "application wave" of domestic large models, they always face cautious supervision from government departments.
Du Guangda, Deputy Director of the Network Security Administration of the Ministry of Industry and Information Technology, introduced at the forum that recently, the Ministry of Industry and Information Technology has launched the special action of "data security escort" in 2023, focusing on key e-commerce platforms, data centers and cloud infrastructure operators to conduct risk troubleshooting, prevention and disposal, and timely resolve the potential risks of data security in the telecommunications and Internet fields.
This forum is hosted by the Shanghai Municipal Administration of Communications and others, and hosted by the Shanghai Cybersecurity Industry Innovation Research Institute and others.
