The U.S. and China are exploring their first meaningful dialogue on artificial intelligence safety since 2024, driven by concerns over frontier AI risks and the potential for catastrophic misuse. Following President Trump's recent state visit to Beijing, Treasury Secretary Scott Bessent indicated that the two AI superpowers would "start talking" and establish protocols to prevent non-state actors from accessing advanced AI models. This marks a significant shift from the Trump administration's initial hands-off approach to AI regulation.

What Changed the Trump Administration's Mind on AI Regulation?

When Trump took office in 2025, his administration opposed restrictions on frontier AI labs, arguing that the U.S. needed to unleash its AI industry to compete with China. That calculus shifted in April 2026 when Anthropic announced Mythos, a cybersecurity model so destabilizing that the company refused to release it publicly. The revelation prompted administration officials to reconsider their stance, culminating in an Executive Order that introduced 30-day voluntary pre-deployment oversight for advanced AI systems.

According to Stanford's 2026 AI Index, U.S. companies still maintain a narrow lead in frontier AI capabilities and dominate advanced chip production. However, Chinese AI firms are closing the gap through open-weight model releases like DeepSeek and are arguably ahead in industrial deployment across the Global South.

Where Could U.S.-China AI Talks Actually Succeed?

Skeptics point to China's history of using diplomatic engagement as cover for other objectives, and a previous round of AI talks in Geneva in 2024 produced no meaningful progress. Verification challenges also complicate any potential agreement. However, experts identify one domain where genuine consensus might emerge: biosecurity.

Last week, chief executives from major AI firms released a letter warning that frontier AI models could erode knowledge barriers to biological weapons development. Since both the U.S. and China have strong incentives to prevent non-state actors from using AI to manufacture bioweapons, there is a realistic possibility that the two countries could agree on screening synthetic DNA or establishing model safeguards. Even a limited agreement would represent an important first step.

"The two AI superpowers are going to start talking and will set up a protocol to determine best practices for AI to make sure nonstate actors don't get ahold of these models," observed Scott Bessent, U.S. Treasury Secretary.

Scott Bessent, U.S. Treasury Secretary

How to Think About AI Governance Across Three Levels

• Frontier Risks: Agentic AI capabilities, AI-enabled biosecurity threats, and scenarios where increasingly powerful models could help bad actors develop dangerous biological agents or other harmful technologies. This is where U.S.-China safety dialogue is understandably focused.
• Governance Questions: Data rules, privacy, user consent, copyright, liability, model transparency, and platform responsibility. Legal and regulatory systems are struggling to keep pace with the technology's rapid evolution.
• Societal Impact: Labor displacement, social mobility, mental health, inequality, human relationships, and institutional reform. This broader layer may prove most consequential over the long run but has seen far less concrete governance experimentation.

What Can China's Labor Courts Teach Us About AI Governance?

While U.S.-China talks focus on frontier risks, China is quietly experimenting with AI governance at the societal level through labor law. A recent court ruling from Hangzhou found that a company illegally terminated a quality-control supervisor for a large language model after claiming AI could perform his work. The court ordered compensation and rejected the argument that AI replacement was justified.

In a similar Beijing case, an arbitration commission ruled that AI adoption constitutes a voluntary business decision, not an unforeseeable external shock. Together, these rulings suggest an emerging legal framework: companies cannot simply fire employees because AI is more cost-efficient. Chinese legal counsel report that employers are increasingly expected to consider reassignment, retraining, and additional compensation if they want to demonstrate good-faith negotiation.

The economic stakes are substantial for China. The AI boom is widening the gap between high-tech strength and broader macroeconomic weakness. While advanced manufacturing, robotics, and AI sectors gain momentum, labor-intensive traditional sectors face strain. The youth job market is already fragile, and AI replacement threatens to aggravate China's structural problem of strong supply but weak demand.

These labor rulings reveal a governance challenge that extends beyond bilateral talks: how to ensure that AI's productivity gains benefit society broadly, not just capital and technology owners. As frontier AI risks dominate headlines, the messier question of how AI reshapes work and inequality may ultimately prove more consequential for both nations.