Early Winners of China's New AI Regulations

The new regulations on AI services are out for review. They seem pretty well-designed for China, but look burdensome, and will create winners and losers.

As Chinese companies have rushed head-first in a mad dash to boost their share prices by announcing upcoming chatGPT-like services, the regulators haven’t just been silently observing. China has come out with a first draft of regulations for this burgeoning industry, and they will play a key role in shaping development.

Others will write a lot more and a lot better about the deeper implications of these new rules, but I’d like to speculate a little bit about what the new rules will mean on a practical level for Chinese tech companies and the industry in general. We can at least attempt to sketch out some early potential winners from these new regulations.

Firstly, like most of China’s internet regulations they place a pretty heavy burden on the companies. Conforming to China’s internet regulations is already burdensome for companies, with ever-changing censorship, the need to strengthen socialist core values, and a whole slew of privacy protections to battle with. It’s not surprising then that these also form the basis of the new AI regulations. They should:

reflect the core values of socialism, and must not contain subversion of state power, overthrow of the socialist system, incitement to split the country, undermine national unity, promote terrorism, extremism, and promote ethnic hatred and ethnic discrimination , violence, obscene and pornographic information, false information, and content that may disrupt economic and social order

In a somewhat surprising move the regulators also demonstrated a very decent grasp of the technology in regulating not just the outputs, but also the processing and manual labeling of data for training. This is not something I saw coming, but it is definitely something that will be impactful. Especially since they seem to want to push some liability for this onto the companies.

For companies this means that compliance for any large model isn’t going to be cheap. This is the type of environment that will favor industry incumbents who are used to dealing with internet compliance, and have the money to set up large and stringent processes around their data and development.

Tencent, and Sensetime come to mind as strong players here. Baidu, historically a heavy-hitter in the AI space, also comes to mind, but their star has been waning for a while now and it’s unclear if this will be the change that allows them to pivot back to greatness.

Adhering to socialist core values, ensuring sufficient diversity in data, ensuring accuracy and truthfulness in a setting where such concepts can sometimes be more fluid than elsewhere, are also somewhat vague targets. These are not easy tasks, especially for large and open-ended models that can generate unpredictable results.

This means that companies that develop AI for specific and limited applications will have an advantage. They will face less uncertainty and difficulty in complying with the rules, and they will be able to develop their products faster, at least in the beginning. For example, instead of creating large and open models like chatGPT, they can use the technology for narrow solutions, like a copilot for excel.

Here, Alibaba seems like a clear winner. Most of their business is in areas like this, and their product portfolio is wide enough to allow for many interesting tie-ins. It’s also the space where it seems most likely we’ll see some startup activity, since the barrier is relatively smaller. They’d still need a lot of backing to make sure they’re in compliance, but development pace is normally very important for startups, especially in China, and this is where you’re most likely to find it.

The new AI regulations in China will have a significant impact on the industry and its players. The rules will likely favor large and established companies that have the resources and experience to comply with the complex and vague requirements. They are also likely to encourage more focused and specific applications of AI, rather than general and open-ended ones. The future of AI in China will depend on how well the companies can adapt to the new environment and how the regulators can balance the goals of security, ethics and development.

An open question is also if AI solutions internationally will favor openness, or if the ability to navigate complex regulations, no-go areas, and constantly shifting sands will become a comparative advantage.