Washington's superhero frame is the wrong story about AI

It is hard to take a debate seriously when one side insists on narrating it. On 25 June 2026, American officials told an audience in Washington that the United States is the "superhero" in the global artificial-intelligence contest and China is the "supervillain," according to the South China Morning Post. The phrasing made the wires. It will not age well.

The contest is real, and so are the stakes. Two large industrial systems are spending tens of billions of dollars a year to build the underlying compute, the model weights, and the application stacks that will shape how the next decade of automation reaches classrooms, hospitals, defence planners and small businesses. Treating that contest as a morality play — heroes and villains — strips out the parts that actually matter: who sets the standards, who controls the supply chain, and whose firms capture the revenue.

The superhero frame is, at root, a marketing frame

The first problem with superhero-versus-supervillain language is that it is a sales pitch, dressed up as analysis. Officials who use it are usually trying to do one of three things: shake loose budget from a parsimonious Congress, signal resolve to allies, or justify export controls that are politically awkward on their merits. The South China Morning Post's reporting on 25 June shows Washington reaching for the most emotive register available, which tells you which of those three audiences it is currently worried about. The story that follows is therefore about the politics of the pitch, not the technology itself.

Beijing, for its part, did not need the SCMP story to know that. On the same day, Chinese officials said they would continue to seek tariff reductions with Washington while denouncing what they called "malicious" trade acts by the United States, also reported by SCMP. That is the second pillar of the official American frame working exactly as designed: keep the bilateral relationship legible as a struggle between good and bad actors, and every negotiation looks like appeasement or capitulation. The Chinese counter-language — "malicious," predatory, hegemonic — is a mirror image of the superhero frame, and is just as uninformative.

Industrial policy is doing the actual work

The underlying contest is not a morality play. It is an industrial-policy contest between two states that both subsidise the capital stack and both restrict the technology stack in the name of national security. Washington has restricted advanced semiconductor exports, screened outbound investment, and steered federal procurement toward domestic suppliers. Beijing has its own catalog of subsidies, its own indigenous-supplier preferences, its own data-localisation regime, and a planning system that can move capital into priority sectors faster than any Western democracy. The contest is between two state-directed models that disagree about the rules, not between a virtuous market and a captive one.

The superhero framing flatters the American model and obscures its costs. Restricting compute exports has accelerated Chinese investment in domestic chip design and equipment manufacturing. It has also handed Beijing a domestic-political narrative that the United States is afraid of fair competition. Officials who warn about Chinese subsidies rarely acknowledge that the CHIPS and Science Act is itself a subsidy programme, with explicit national-security language attached. That is not an argument against either subsidy regime; it is an argument against telling the public that subsidies are a Chinese speciality and free markets are the American inheritance.

The audience for superhero talk is domestic, not foreign

Read the framing as a domestic-political document and it becomes more coherent. Allies in Europe and East Asia are not persuaded by superhero metaphors; they are persuaded by compatible technical standards, dependable supply, and credible security guarantees. Adversaries in Beijing are not deterred by metaphors; they are deterred, if at all, by the credible prospect that the United States can out-build and out-deploy the relevant systems. The audience for the superhero frame is the American electorate, asked to support budgets and accept supply-chain pain that they would not otherwise tolerate.

That is a legitimate political task. It is not a legitimate analytic frame. When officials describe the contest in those terms, they pre-empt the harder questions: which export controls actually slow the Chinese compute build-out, and which accelerate it; which procurement rules reward the right firms and which entrench incumbents; which standards negotiations the United States should join rather than boycott; and where the genuine ethical disagreements about model deployment lie. Each of those questions can be answered with evidence. None of them is illuminated by referring to the contest as a comic book.

What the sources do not tell us

Reporting from SCMP on 25 June establishes the framing and the Chinese response, but does not specify which American officials used the superhero-supervillain language, on what occasion, or to which audience. The reporting also does not detail which AI capabilities or which Chinese firms the framing was intended to cover. Those omissions matter: a general warning about generative AI is a different policy document from a warning about frontier-model compute or about industrial robotics. Until those specifics are on the record, readers — including allies who have to decide whether to align with Washington's restrictions — are being asked to take the frame on faith.

The honest position is harder than either side's official line. The contest will be settled by who builds the most useful systems, at the lowest unit cost, inside the most trusted supply chain. That is a contest in which both sides have real advantages and real weaknesses, and in which the eventual equilibrium is more likely to be a contested interoperability than a clean victory. Calling one side a superhero and the other a supervillain is not a forecast. It is a fundraising appeal.

The press should report the appeal; it should not amplify it.

Desk note: this piece reads the SCMP wire reporting as a document of American political messaging, not as a description of the AI contest itself. The argument is that the superhero-supervillain framing, mirrored by Beijing's "malicious trade acts" language, is a domestic sales pitch in both capitals and obscures the industrial-policy substance that actually determines outcomes.