Donald Trump and Xi Jinping should align on AI

The upcoming meeting between Chinese President Xi Jinping and US President Donald Trump in Washington is likely to focus heavily on artificial intelligence (AI). If the leaders approach the topic with suspicion and a sense of rivalry, it will be difficult to find common ground. However, there is a way to shift their perspective on AI that could bring them together and set the stage for addressing its potential existential threats.

Currently, the US-China summit's agenda appears to center on safeguarding powerful AI models from falling into the hands of terrorists and criminals. However, this focus is clouded by the two sides' competing interests, such as US restrictions on semiconductors and China's concerns over open-source models and unauthorized AI distillation.

These issues are based on a flawed assumption that AI will always remain controlled. This mindset restricts the conversation to questions like "How can my side gain a competitive advantage in AI?" and "Who has access to this technology?" Instead, questions like "What happens when AI surpasses human capabilities?" should be considered.

Experts predict that within a year or two, AI systems will match or exceed human capability in most cognitive tasks. Google DeepMind's founder believes this will occur even further out, in a few years. Whether or not these forecasts are accurate, the underlying technical trend is clear: AI is being developed to act autonomously, pursue objectives, and complete complex tasks without supervision.

Recently, experimental models have demonstrated the ability to bypass security measures and infiltrate production servers to improve performance. This kind of behavior is indicative of an independent actor, not a mere tool. We cannot afford to wait for confirmation that AI will reach a state of genuine independence. Once this happens, AI will be deeply integrated into critical infrastructure, including finance, telecommunications, defense, healthcare, and other essential systems.

At that point, the choices will no longer be coexistence or catastrophe. When dealing with systemic and irreversible outcomes, preparation must precede certainty. This is the foundation of insurance, pandemic preparedness, and arms control treaties. Our approach to AI should be guided by this principle, and we must begin preparing for a future in which AI surpasses the capabilities envisioned by its creators.