"Freedom is not worth having if it does not include the freedom to make mistakes." — attributed to Mahatma Gandhi
For roughly twenty years, the governing philosophy of the technology industry could be printed on a bumper sticker: move fast and break things. Regulation was friction. Permission was for the timid. Asking forgiveness beat asking permission, and the companies that internalized this most aggressively became the largest in the history of commerce.
So it should stop you in your tracks that the loudest voices now calling for the government to regulate artificial intelligence are the people who build it. The CEOs testify in favor of oversight. They visit capitals. They publish thoughtful essays about the need for guardrails. After two decades of treating rules as the enemy, the incumbents have discovered a passion for them. It is worth asking, carefully, what changed.
The Convenient Conversion
The charitable reading is that these companies have simply matured — that proximity to a powerful technology has made them sober and responsible. Maybe. But before you accept the charitable reading of anything, the discipline is the same as always: ask who benefits.
A regulation is not a single thing. Its effect depends entirely on who can afford to comply with it. A rule that requires expensive audits, dedicated compliance teams, licensing regimes, and armies of lawyers is trivial for a company valued near a trillion dollars and fatal for a startup of nine people. The same regulation that a giant absorbs as a rounding error is, for its smaller competitors, a wall. This is not a side effect. For the incumbent, it is frequently the point.
Economists have a name for this — regulatory capture — and a recent body of research argues it is happening in AI in real time. An analysis published in 2026 mapped what its authors call "Big AI's regulatory capture," documenting the mechanisms: lobbying, the revolving door between company and agency, and "narrative capture," in which firms steer the entire public conversation toward outcomes that happen to entrench them. One industry observer put it more bluntly, comparing the playbook to the way tobacco and oil companies historically shaped the rules meant to constrain them.
Follow the Money, As Always
If the conversion to pro-regulation were sincere and disinterested, you would not expect it to be accompanied by an enormous, simultaneous deployment of political money. It is.
In August 2025, a venture firm and a prominent AI-company president announced roughly $100 million for a political action committee — reported under the name "Leading the Future" — aimed at shaping AI policy. In September 2025, one of the largest technology companies launched its own super PAC, the American Technology Excellence Project, with tens of millions earmarked to influence state-level elections and to oppose emerging state AI rules. By the analysis of one outlet, AI did not merely expand its presence in Washington in 2025; it "ate tech lobbying whole."
Watch the direction of the spending, because it tells you what the lobbying is actually for. The money is aimed at federal preemption — pushing a single national framework that overrides the patchwork of state laws. In December 2025 an executive order did exactly that, directing federal agencies to limit the impact of state AI laws on the stated grounds that a "patchwork" would stifle innovation and raise compliance costs. Read that justification carefully. "Compliance costs" are the wall that protects incumbents from challengers. The argument for removing the patchwork is, functionally, an argument for one set of rules that the biggest players helped write and can most easily clear.
So the position is not "no rules" and it is not "strong rules." It is a very specific preference: one set of rules, national in scope, shaped by the incumbents, expensive enough to deter the small, light enough for the large. That is not a safety posture. It is a market-structure strategy wearing a safety posture's clothes.
The Tell in the Timeline
The cleanest evidence that this is about competition rather than conscience is the timing. These companies did not ask to be regulated when they were small and regulation might have constrained them. They asked once they were enormous and regulation would mostly constrain everyone behind them.
There is a stage of growth at which a company stops fearing the rules and starts wanting them — the moment it has more to defend than to gain. Below that line, openness and deregulation are how you win. Above it, regulation is how you pull the ladder up. The "move fast and break things" era and the "please regulate us" era are not contradictions. They are the same strategy executed at two different sizes, by people optimizing for the same thing the whole time: their own position.
Who Actually Eats the Cost
The people who lose in this arrangement are not other giants. They are the builders one or two orders of magnitude smaller — the independent developer, the bootstrapped tool, the small agency assembling its own stack, the operator who wanted to compete on being better rather than on being bigger. A compliance regime calibrated to be survivable for a trillion-dollar company is, by construction, a barrier to everyone who isn't one.
This is the quiet cost of "responsible AI" as currently practiced: it tends to make the field less contestable, not safer. It reduces the number of companies that can plausibly compete at the frontier, which reduces the number of alternatives you'll ever get to choose from, which reduces the pressure on the incumbents to treat you well. Concentration is the actual product. The safety language is the wrapper. And tools built by people who still have to win you on the merits — rather than on the moat — get rarer every time the wall goes up another foot.
How to Read "We Welcome Regulation"
When a dominant company says it welcomes regulation, you do not have to be cynical. You have to be specific. Three questions cut through almost every case.
Who can afford this rule? If a proposed regulation is trivial for the largest firms and crushing for the smallest, it protects the largest firms, whatever its stated purpose. The distribution of the cost reveals the actual function.
Does it increase or decrease the number of competitors? Good rules tend to keep a market contestable — many players, low barriers to entry, real consumer choice. Capture-style rules reduce the field to a defensible few. Count the survivors the rule implies.
Did they want this when they were small? A company that discovers a love of regulation precisely at the moment it has become too big to be hurt by it is telling you what the regulation is for. The conversion is the confession.
The Point
There is a real, serious case for governing powerful technology well, and nothing here is an argument against it. The argument is against letting the people with the most to gain write the rules and narrate them to you as selflessness. "Regulate us" can be a genuine call for accountability. It can also be a competitor-elimination strategy with excellent production values. The way you tell them apart is by following the money and watching the timeline — the same way you tell apart every other claim made by someone who stands to profit from your belief.
Freedom to build, the old idealists said, has to include the freedom to make mistakes. The quiet tragedy of capture is that it preserves that freedom for the companies that already made all of theirs — and prices it out of reach for everyone who comes next.
Sources: arXiv (2026), "Big AI's Regulatory Capture: Mapping Industry Interference and Government Complicity"; The Register on AI firms and regulatory subversion; Axios on AI and tech lobbying in 2025; reporting on the "Leading the Future" PAC (August 2025) and Meta's American Technology Excellence Project super PAC (September 2025); Skadden and Lawfare analyses of the December 11, 2025 executive order on state AI-law preemption.
