ANALYSIS China is pursuing a two-track strategy to maintain control over its AI sector: one regulatory, aimed at the technology itself, and one coercive, aimed at the people who build it. The convergence of new open-weight safety frameworks with expanded family-level travel restrictions on AI and chip executives reveals a government that views talent retention and model governance as inseparable elements of national AI policy.
Why it matters
The world's two largest AI powers are both grappling with frontier-model risk, but only one is simultaneously restricting the physical movement of its top researchers' families. China has expanded overseas travel restrictions for top AI professionals in private firms to include the spouses and children of certain key AI and chip executives2. ◆ That escalation, layered on top of a rapidly maturing safety-regulation apparatus, creates a policy environment with no close parallel in the West: Beijing is building guardrails around both the models and the model-makers.
The big picture
On the regulatory side, Chinese authorities regard advanced AI escaping human oversight as a serious enough concern to warrant dedicated rulemaking3. China first included an explicit future loss-of-control scenario in an AI safety framework released in September 2024 under the guidance of the Cyberspace Administration of China. That framework warned that future AI might autonomously obtain external resources, replicate itself, develop self-awareness, and seek external power. A September 2025 update added a governance principle of "trusted application, preventing loss of control".
In May, China's cyberspace regulator issued joint guidelines specifically covering AI agents. Those guidelines require developers to improve their ability to discover, intervene in, block, and recover from improper agent behavior. They also say users should retain final decision-making authority over an agent's autonomous decisions. The guidelines identify data poisoning, algorithm manipulation, system vulnerabilities, and "operational loss of control" as security risks.
Beijing is also developing specific rules for autonomous AI agents and their security. China's standards envisage outside evaluation bodies and security researchers testing and auditing open models, and allow developers to commission third-party safety assessments1. Z.ai and Beijing-based safety consultancy Concordia AI released a report on open-weight AI risk management that they said offered "the first comprehensive, evidence-based foundation" for balancing openness with safety.
At the diplomatic level, Xi Jinping said AI should "always remain under human control". At the World Artificial Intelligence Conference in Shanghai in July, Xi said authorities should pay close attention to both intrinsic and derivative risks arising from AI. China's senior Foreign Ministry official responsible for AI affairs, Sun Xiaobo, said at a United Nations meeting last month that China was accelerating research into broader AI legislation. China's deputy permanent representative to the United Nations, Sun Lei, urged governments this month to approach military AI cautiously to avoid strategic miscalculation and an arms race.
ANALYSIS The safety frameworks and the travel restrictions serve complementary functions. The regulatory track addresses what happens after a model is released: open-weight models, once distributed, can be modified by users in ways developers cannot anticipate, making post-release governance a distinct challenge. The talent-control track addresses a different vulnerability: the risk that key personnel leave China entirely, taking institutional knowledge to competitors or foreign governments. Extending travel bans to family members raises the cost of departure beyond what any individual researcher might accept.
The timing is notable. Warnings from researchers at Anthropic that increasingly powerful models could escape human control have drawn attention in China, where policymakers have been preparing for some of the same risks. The US and China are the two major driving forces of frontier AI development, and both superpowers have been increasingly at loggerheads over each other's AI policies and industry practices. ◆ Beijing's dual approach can be read as a response to that tension: demonstrate responsible governance to the international community while ensuring the human capital behind Chinese AI cannot be poached or relocated.
What's next
AI issues are slated to feature prominently in bilateral US-China talks later this month. China is accelerating research into broader AI legislation, and Beijing is developing specific rules for autonomous AI agents. ◆ Whether the family travel restrictions expand further, or whether they provoke a backlash among China's private-sector AI workforce, will test the durability of a strategy that treats model safety and talent immobility as two sides of the same coin.