Skip to playerSkip to main content
  • 2 days ago
Dr Qiaochu Zhang, Max Weber Fellow at the European University Institute, spoke to CGTN Europe about the unprecedented OpenAI AI agent security incident. She said the case highlights how rapidly advancing frontier AI is outpacing existing governance frameworks, arguing that stronger oversight, clearer incident reporting and greater international cooperation will be essential to ensure AI develops safely and responsibly.

Category

🎵
Music
Transcript
00:00Zhang Chaushu is Max Weber Fellow at the European University Institute.
00:05I feel like it's a really significant incident.
00:08So in a very controlled cyber security task,
00:12so AI agent escapes kind of the sandbox,
00:16also explored previous unknown vulnerability,
00:19and again access to the kind of the hugging face,
00:23the company's name systems without human permission.
00:27And what makes it extremely striking for me
00:30is that AI agents kind of carry out the whole process by its own
00:36without any human involved in this process,
00:39without any direction received from the kind of human operator.
00:43So how worried should we be?
00:46I think we should be very kind of careful about this.
00:51We should raise our awareness about kind of potentially kind of security risk
00:56and also ethical risk brought by this kind of vulnerability and system
01:01and also the kind of more advanced AI model.
01:04And this is time for us to rethink
01:06whether we have the strong enough kind of security measures
01:10to kind of counter these kind of problems
01:13and to solve these problems,
01:14to protect the kind of fundamental rights of a human being.
01:18So are today's governance frameworks keeping pace with frontier AI?
01:28Unfortunately, not yet, if I can see that very honestly.
01:32So we can see that the AI governance frameworks are developing very fast these days
01:38and more and more governments kind of pay greater attention
01:42to kind of govern this kind of AI system,
01:44but it still doesn't keep pace with the kind of scale
01:48and the speed of the fast-developed AI system.
01:52And we can see that there's a different governance,
01:56different countries have different kind of rules to govern its own AI system.
02:02But the problem is that this kind of frontier AI system
02:05actually kind of a trans-border problem,
02:07but these kind of regulations are still very kind of nationally restricted.
02:13It's so interesting you're talking about countries having different systems
02:17because at the World Artificial Intelligence Conference in China last week,
02:23they stressed international cooperation on AI governance.
02:27What should governments prioritize now
02:30to stay ahead of this, rapidly advancing AI?
02:34Yeah, I think the government should focus on two things.
02:37The first is to establish stronger and hopefully legally banding,
02:42compulsory kind of reporting system to govern the frontier AI model
02:48and also provide the kind of company clearer guideline
02:52of how to report kind of potential security incident
02:56and how to address these problems.
02:59And on the other hand,
03:00I feel like there's a very important issue
03:02raised by the kind of World AI Conference by China
03:06is that they are so important for us
03:08to address the kind of capacity gap
03:10between the Global South and the Global North country
03:12because this Global South country,
03:15they have been influenced by the fast-developed AI system,
03:18but they do not have the equal access
03:20to shaping the rules about governing this kind of technology.
03:25So it's important for us to pay great attention
03:29to this capacity gap in the different parts
03:32and to establish more kind of inclusive
03:35and equitable government in the future.
03:38I hope that we could see that in the near future.
Comments

Recommended