Four things to know about China’s new AI rules in 2024 ARTIFICIAL INTELLIGENCE Four things to know about China’s new AI rules in 2024 From drafting an AI law to analyzing copyright and security, experts say these are the priorities for Chinese AI regulators in 2024.
by MIT Technology Review February 7, 2024 Generate PDF Download Last year was remarkable for Artificial Intelligence. Thanks to products like ChatGPT, many millions of people now directly interact with AI, talk about it, and deal with its impact every day.
Some of these people are policymakers, who have strived to respond to the problems that AI products pose without reducing our ability to take advantage of what it has to offer.
So earlier this year, my colleagues and I looked around the world for signs of how AI regulations are likely to change this year. We summarize what we found here.
In China, one of the key movements to watch in 2024 is whether the country will follow in the European Union’s footsteps and announce its own comprehensive Artificial Intelligence law. In June last year, the Asian country’s main government body released a list of legislation it was working on. An “Artificial Intelligence Law” appeared for the first time.
The Chinese government is already good at reacting quickly to new technologies. The region was probably the first in the world to introduce legislation on generative AI just a few months after the huge success of ChatGPT. But a sweeping new law could give China even more control over how AI impacts (or doesn’t impact) the way things work today.
But you shouldn’t just take my word for it. I asked several experts on Chinese AI regulations what they think will happen in 2024. So I’ll share the top four things they said to expect this year.
Don’t expect the Chinese “AI Law” to be finalized anytime soon
Unlike previous Chinese regulations, which focus on subsets of AI such as deepfakes, this new law is aimed at the big picture, which means it will take a long time to draft. Graham Webster, a research scholar at Stanford University’s Center for International Security and Cooperation, hypothesizes that we’re likely to see a draft AI Act in 2024, “but it’s unlikely to be finalized or come into force.”
A big challenge is that even judging what is and isn’t AI can be so complicated that trying to resolve everything with just one law can be impractical. “It is always a question of law and technology whether a single law is necessary or whether it should be approached in terms of its applications in other areas,” says Jeremy Daum, who researches Chinese laws at the Paul Tsai China Center. “So a generative AI content regulation makes sense, but only for AI? Let’s see what happens.”
China’s government is telling AI companies what they should avoid
In 2023, the Chinese Academy of Social Sciences, a state-run research institute, drafted a consultative version of the future AI law, and it is a useful reference for what China wants to achieve. One of the most interesting items in the document is a “negative list” of existing areas and products that AI companies should stay away from unless they have explicit government approval. It will be interesting to see what will be included in this list and how it differs from similar bans set out by the EU.
“The list subjects only certain products, services and model development to rigorous oversight and was created with the intention of reducing the overall regulatory compliance burden on Chinese companies,” says Kristy Loke, a research fellow at the Center for the Governance of AI, a think tank. The list tells companies exactly where they shouldn’t go to stay in the government’s good graces, which should help them avoid accidentally angering Beijing.
Third parties can start evaluating AI models
Regulations mean nothing if they are not enforced. So developing a way to evaluate AI models could be on Chinese regulators’ checklist for 2024, says Jeffrey Ding, assistant professor of political science at George Washington University.
How could this be done? “First, developing a national platform to test and verify the safety and security of models, and second, supporting third-party assessment organizations to implement regular reviews,” says Ding.
(In this regard, I wrote about a fascinating and highly detailed paper released by Chinese technology companies and academics last year that suggested ways to evaluate AI models.
China is likely to be lenient on copyright
Generative AI has created a copyright nightmare, and current laws are not up to the task of clarifying who owes what to whom and why. Angela Zhang, a law professor at the University of Hong Kong, expects more policy guidelines and court rulings from China to clarify potential intellectual property issues next year.
China’s government will likely be lenient on AI companies. “Given the overarching national agenda to encourage the growth and development of the AI sector, it is very unlikely that Chinese administrative agencies will adopt an aggressive stance in investigating companies for AI-related infractions. Meanwhile, Chinese courts will adopt a business-friendly approach in deciding intellectual property cases,” says Zhang.
Needless to say, I will be keeping an eye on all four areas and will update you on them.
( fonte: MIT Technology Review )
