BEIJING – Amid signs of stabilising ties, Beijing and Washington are taking tentative steps to cooperate on artificial intelligence (AI), with analysts suggesting that military applications and transparency around AI-generated content could be high on the agenda.

But tempering optimism over the collaboration is the underlying distrust and competition between the two tech powerhouses, which pose a potential obstacle to progress, said observers.

In a Bangkok meeting on Jan 27, United States National Security Adviser Jake Sullivan and Chinese Foreign Minister Wang Yi discussed steps towards holding an AI dialogue in spring, although neither the date nor the specific agenda has been announced so far.

Their dialogue was one of the main outcomes of a top-level meeting between US President Joe Biden and Chinese President Xi Jinping in San Francisco in November 2023, which stabilised bilateral relations after a turbulent year.

The director of geo-technology at political risk consultancy Eurasia Group, Ms Lu Xiaomeng, believes the US and China are working towards an understanding of key concepts, such as the situations in which human control over machines should be retained.

Reports during the Xi-Biden summit said that the Americans were seeking Chinese agreement on keeping AI out of the command and control of nuclear arsenals, but no concrete outcome was reached on this specific issue.

Both sides are also engaging in their own policy debates on how to improve the transparency of generative AI content, said Ms Lu, referring to algorithms that can produce text, images and audio from human instructions, made popular by the public release of ChatGPT in November 2022. “Bilateral recognition of various content transparency mechanisms could be beneficial to society,” she said.

But Dr Ilaria Carrozza, a senior researcher at the Peace Research Institute Oslo, warned that both countries are modernising their militaries, and view AI as playing a big role in the upgrading of their armed forces.

“Agreeing on a common language which stops short of a ban (on such use) is unlikely to have any substantial impact on this broader trend,” she said.

Technology remains at the heart of the competition between Beijing and Washington, said Dr Carrozza, especially regarding dual-use AI technology for both military and civilian applications.

“(Hence) there are still many areas where finding an agreement will be difficult if not impossible,” said Dr Carrozza, whose research specialisations include China’s foreign policy and AI.

“At the same time, issues of safety and risks associated with the use of AI technologies are a concern for both American and Chinese policymakers; those are potential areas where we might see some progress.”

The overall distrust in the bilateral relationship could pose an obstacle. The US has worked with other countries since late 2022 to restrict exports of computer chip technologies to China, which some see as an effort to limit its AI development.

China has been the less direct of the two in articulating what it wants out of bilateral AI cooperation.

Experts in China have called for collaboration in AI research and governance, noting that both countries share basic beliefs in ethical use, risk assessment and expanding international cooperation.

In a January 2024 commentary, Professor Sun Chenghao and Professor Liu Yuan from Tsinghua University’s Centre for International Security and Strategy called for starting US-China exchanges on the basic principles of AI development, ethical concepts and other related terms.

The Tsinghua University centre and the Washington-based Brookings Institution think-tank have held regular Track II dialogues, which involve non-official representatives, to discuss the impact of AI on international security since 2019.

Global cooperation on AI remains in its infancy, even as the technology has advanced, and harmful applications are already widespread in the form of deepfakes, scams and misinformation.

The US has led efforts to build safeguards around military AI, with 52 states so far – including Singapore – endorsing a non-binding Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy launched in February 2023. But China is not part of the grouping.

China was among 29 signatories of the Bletchley Declaration in November 2023, announced after the first global summit on AI safety held in Britain, where countries agreed to a new joint effort to ensure AI is developed and deployed in a safe and responsible way.

The European Union in December 2023 agreed on a landmark AI regulation deal, which includes governments’ use of AI in biometric surveillance and transparency requirements for AI models.

What will US-China AI cooperation mean for the rest of the world?

Political scientist Zhang Junhua said that unlike climate change, AI is much more sensitive because of its dual use. Thus, AI cooperation will not be as broad as that for climate change, where the two countries’ cooperation was crucial in galvanising global climate action, such as the groundbreaking Paris Agreement of 2015.

He added that China is aware that despite its immense progress in AI research and development, particularly in surveillance-related applications, the US is still better in fundamental research. The latter refers to research that may not have immediate applications, but forms the foundation for future breakthroughs.

“This is the driving force for China in seeking cooperation. But for exactly this reason, the US has become extremely cautious,” said Dr Zhang, a senior associate at the Brussels-based European Institute for Asian Studies, a policy and research think-tank.

Dr Carrozza believes many countries in the Global South are hoping that tensions will not jeopardise cooperation.

The absence of a shared AI governance framework can lead to multiple strategies emerging and being adopted by different states or groups of states, and that may even have different goals, she said.

“We can already see that three normative spheres have emerged, championed by the US, the European Union and China, each with a different approach to governing AI, which leaves other countries in the difficult position of having to choose.”