Two rivals, one AI policy: gates, not laws

Under rules that surfaced in late September, the spouses and children of senior AI engineers at Chinese companies like Alibaba and DeepSeek now need government approval before any trip abroad. Beijing has long controlled the travel of its state-employed scientists; the new order extends the same exit control to the families of private-sector talent, and states the purpose plainly — to keep technical know-how from reaching the United States ^[1]. The same weeks produced the rhyme on the other side of the Pacific. On September 9, the American labs — OpenAI among them — wrote to Congress asking for mandatory national safety rules ^[2]. The letter's case was that the technology is now moving faster than promises can cover.

The prospect of AI-accelerated AI development demands more than voluntary commitments. — OpenAI

What Washington sent back, this month, was the opposite of a statute: a voluntary accord signed in the White House, and a Super Intelligence Force chaired by the director of national intelligence, Jay Clayton, whose first deliverable is a report due in 120 days ^[3]. Washington's version of control has been assembling all year, and it points at everything except a binding rule. After Anthropic's Mythos model penetrated classified government systems in a test called Project Glasswing — in hours, not weeks — the administration's answer was a voluntary 30-day pre-release review that gives national security agencies early access to frontier models, signed by OpenAI, Google DeepMind, Microsoft, xAI and Anthropic ^[4]. The same incident produced a different kind of rule: the administration suspended foreign nationals' access to Anthropic's Fable and Mythos models ^[5]. And in January, Trump signed an executive order asserting federal control over AI regulation and directing agencies to weigh cutting grants to states that pass their own AI laws — explicitly so controls could not put American firms at a disadvantage against China ^[6]. Beijing's moves are the same instruments, pointed the other way. Washington runs its controls inbound — who may log in — and downward, against its own states; Beijing runs them outbound, on people. Where Washington barred access to a model, Beijing drafted restrictions on overseas access to its own most advanced models, weighing limits on API access, bans on model-weight downloads, and making the leak of AI technology a national security offense ^[5]. Where Washington preempted its own states, Beijing ordered Meta to unwind its $2 billion purchase of the startup Manus and barred two of Manus's co-founders from leaving the country — dismissing the firm's relocation to Singapore as "Singapore-washing," and now requiring sign-off before ByteDance, Moonshot or StepFun accept American capital ^[7]. The one thing the two capitals did together was refuse. In late September at the UN General Assembly, Trump ordered the term "artificial intelligence" replaced by "super intelligence" across federal documents and called international regulation a "globalist scheme." That same week, the leaders of OpenAI and Anthropic briefed the UN Security Council asking for the global oversight their governments were declining. Dario Amodei made the ask plain.

I believe that this is the most important global security issue facing the world today. — Dario Amodei

The two superpowers jointly rejected the international rules those executives were requesting ^[8]. It was the only superpower cooperation on the record this month — a shared no. Each capital states its reason in its own words, and it is the same sentence with the flags swapped. At the United Nations, Trump framed any binding rule as a hoax and a gift to Beijing.

We will not in any way hinder or stifle the Growth of this incredible Industry. Rather, we will cherish it, help it, and watch over it, as it grows! — Donald Trump

Clayton, the new force's chairman, supplied the national-security phrasing.

Not being first increases the identified, and unidentified, risks, particularly from our adversaries. Being first will better enable us to address those risks on behalf of the American people. — Jay Clayton

Beijing's order on the engineers' families carries the same logic in reverse — the control exists to stop know-how from reaching the United States ^[1]. Neither government explains the other's move, and neither needs to; each cites the other as the reason for its own. Back to the family in Beijing, waiting on a travel approval. Every restraint the month's record tightened sits at a border: who may leave, who may log in, which state may pass a law of its own. The labs asked, in writing, for a safety obligation. What both capitals built instead was passport control, applied to AI.