Washington wants a share of the AI boom and reassurance that somebody else has the risks under control. Preferably the same people producing the boom. That is a comfortable political arrangement until those people start discussing how much power they might actually acquire. Then the enthusiasm becomes irritation, and the invitation to help shape the future comes with a request to please stop describing it quite so vividly.
Maria Curi reports in Axios that AI leaders’ alternating promises of prosperity and warnings of catastrophe are alienating Washington. My reading is that the discomfort runs deeper than a communications problem. A political establishment accustomed to treating business leaders as donors, employers and occasional witnesses is confronting people who aspire to control infrastructure through which much of the economy could eventually operate. The awkwardness is about authority. Who gets to make consequential decisions, who can refuse whom, and who will be left explaining the consequences?
The phrase “clipping coupons” fits rather well here. It describes the desire to collect the yield from an asset without doing much of the difficult work of maintaining it. In political terms, the coupons are the prospect of growth, investment, strategic advantage and an attractive story about national renewal. Every administration would like to preside over a technological renaissance. Every politician would like the new facilities, the successful businesses and the announcement at which prosperity appears to have a local address. The troublesome part is acknowledging that the people delivering those things may develop bargaining power that a hearing and a stern letter cannot easily dissolve.
Consider what dependence would mean. If a handful of companies became essential suppliers of advanced reasoning tools, scientific assistance, software production and administrative capacity, their decisions would reach well beyond their own balance sheets. Pricing, access, deployment schedules and restrictions could become matters of public consequence. That is a conditional argument about where concentration can lead, not a claim that AI executives already run the government. But it explains why the promise of extraordinary capability should provoke a question about extraordinary private leverage. The bigger the promised transformation, the less plausible it becomes to treat its owners as ordinary vendors.
The executives, of course, do themselves no favors when they imply that they alone understand both the magnificent opportunity and the terrible danger. There is a convenient structure to that position: trust our judgment because the technology is powerful, and give us greater influence because its power is dangerous. Even sincere concern can produce a self-serving political argument. A warning deserves examination; it does not automatically entitle the person issuing it to write the rules. Nor does technical competence confer a mandate to decide how much disruption everyone else should tolerate. A society is entitled to ask who benefits, who bears the losses and what happens when the experts disagree.
Yet the corresponding political temptation is equally convenient. Welcome the investment, celebrate the potential, invoke the international competition and then act astonished when the companies involved behave like independent centers of power. You cannot consistently advertise a technology as capable of transforming national fortunes while assuming its ownership has no implications for national authority. Well, you can say both things. Making them fit together is harder.
There is another reason this relationship produces so much instability: Washington’s political confidence can exceed its understanding of the technology. My view is that too much of the debate treats AI as an economic miracle, a security weapon or an impending catastrophe without developing the technical understanding needed to distinguish those possibilities. That leaves policymakers vulnerable to whichever promise or warning has most recently caught their attention. When officials cannot independently assess a company’s claims, they can become dependent on its executives for explanations while resenting that very dependence. Fear and deference then feed each other.
Trump’s volatility adds a further source of uncertainty. In my reading, his tendency to turn issues into immediate tests of loyalty, strength and personal advantage makes a consistent approach to AI harder to sustain. A technology with long investment cycles and uncertain consequences requires patient judgment; a presidency driven by abrupt shifts in emphasis can amplify every disagreement into a change of direction. The danger is a government trying to accelerate adoption, assert control, reassure the public and confront the companies all at once, without deciding how those objectives fit together. Businesses and public institutions then have to plan around the possibility that today’s encouragement becomes tomorrow’s confrontation. When government tries to run in opposite directions simultaneously, the result is policy whiplash, and the costs travel well beyond Washington.
This is why the fixation on tone feels inadequate. More polished executives could make the relationship more pleasant without making it more accountable. A reassuring presentation does not establish a right to inspect a system. A cheerful promise does not create an effective route for challenging a decision. A carefully worded commitment does not guarantee that a public institution can change suppliers when it needs to. Those are questions of institutional design, contracts, competition and enforceable obligations. They remain after everybody has received media training.
There are legitimate reasons for Washington to be wary of concentrated AI power, just as there are legitimate reasons to pursue the technology’s benefits. The test is whether that wariness produces workable public safeguards or merely a demand that uncomfortable possibilities disappear from the sales pitch. Government should have enough independent expertise to interrogate claims, enough procurement discipline to avoid unnecessary dependence and enough clarity about responsibility to prevent every failure from becoming somebody else’s problem. Companies should be expected to substantiate what they promise and what they fear. Neither side should get to use uncertainty as a permanent exemption from scrutiny.
The most revealing political anxiety may be the possibility that the coupons come attached to obligations. If AI delivers even a substantial portion of its advocates’ ambitions, elected officials will have to make decisions about the distribution of its gains and the limits of its owners’ discretion. They will not be able to collect the prestige of technological leadership indefinitely while outsourcing every difficult judgment. Washington wants AI leaders powerful enough to deliver prosperity and manageable enough to remain politically convenient. The friction begins when it becomes clear that those wishes do not necessarily travel together.