The sharpest fact in the debate over what Alex Lo calls American AI imperialism is about power, not a future forecast: advanced semiconductors, computing infrastructure and AI capabilities are concentrated mainly in the United States and China, Chatham House said in an Aug. 7 analysis. Many countries outside those powers depend on systems they neither develop nor fully control.
Lo’s Monday column in the South China Morning Post frames that imbalance as American AI imperialism. The label is an opinion, not an established description of Anthropic’s policy. But the question beneath it deserves a plain answer: democracies can seek an AI lead without claiming the right to write rules for everyone else. The United States and its allies should pair restrictions with meaningful international participation.
A democratic lead is a real security argument
In his January 2026 essay, The Adolescence of Technology, Dario Amodei argues that democratic countries face a tension: they want to stay ahead of authoritarian governments while developing AI carefully. He proposes using restrictions on chips and semiconductor equipment to slow authoritarian countries’ progress and give democracies more time. Anthropic made a related case in its May 14 paper, arguing that the United States and its allies should maintain an AI lead and backing tighter controls on advanced computing access to China.
That position has a serious rationale. Amodei warns that advanced AI tools could help authoritarian governments strengthen repression, and Anthropic argues that the political systems leading AI development will shape global norms. Those are arguments about risks and likely consequences, not independently verified findings in the cited policy work. They should be debated, not treated as proof that a U.S.-led lead will prevent abuse.
Amodei’s position is also more complicated than a call for Washington to act alone. He supports safety discussions with Chinese AI experts, and he warns that tools used against autocracies could be turned against democratic societies at home. That acknowledgment matters: a democratic lead is not a safeguard by itself if the same capabilities can threaten liberties domestically.
Leadership cannot substitute for consent
The strongest case for controls is that a government should not ignore the possibility that powerful AI systems could serve repression or security threats. But the strongest case does not settle who gets to define the threat, set the limits or bear the consequences. When policy centers the United States and its allies as the leaders, countries that depend on systems they do not control risk having less influence over the rules that govern them.
Free newsletter
Get the morning briefing
Start each day with the AI stories that matter and why — a short, free email from our newsroom.
That is the imbalance Chatham House describes; it does not establish that every developing country is excluded from AI governance. Nazam Laila of Chatham House argues that international cooperation can help countries outside the U.S.-China competition coordinate, build capacity and strengthen their ability to shape rules. She also says those efforts cannot counter the technological dominance of the two leading powers. A seat at the table will not erase unequal capabilities, but that is a reason to make participation consequential, not to dismiss it.
There is also a practical limit to treating export controls as a complete answer. In a March 25 analysis, SIPRI researcher Kolja Brockmann said controls on AI models, data and algorithms face unclear standards and difficulties classifying and enforcing restrictions on intangible transfers. That does not prove chip controls cannot work. It does mean governments should not present controls as a simple, self-validating guarantee of democratic security.
Make the rules answerable to more countries
The imperialism charge should not erase the security concern Amodei raises. Nor should that concern grant a small group of governments an open-ended mandate to decide the future of a technology on which many others depend. Both points can be true: authoritarian misuse is a serious policy concern, and technological leadership is not the same as global consent.
The U.S. government and its allies should publish the security case for AI-related export restrictions, invite affected governments—including those without major computing capacity—to help shape shared standards, and review restrictions against their stated aims. They should also make participation a regular part of governance decisions, rather than an afterthought once the leading powers have set the terms.
That approach will not equalize AI capabilities or remove the risks Amodei identifies. It would, however, make the exercise of power more accountable to the countries expected to live with its consequences. Democracies can lead. They should not mistake leadership for permission to decide for everyone.