The Corporate Capture Mapping: Who Benefits When the Builders Preach Doom?

When the people building a technology tell you it might end the world — and then ask to write the rules for it — the correct response is not fear. It is the question this desk exists to ask: who benefits from this story?

For two years that question could be dismissed as cynicism. Now it has a dataset. In May 2026, an international research team published a systematic mapping of how "Big AI" influences the laws and oversight meant to govern it. They defined twenty-seven distinct patterns of corporate capture — the documented mechanisms by which regulation ends up serving corporate rather than public interests — and applied the framework to a hundred published accounts surrounding the key regulatory moments of 2023 through 2025, from the EU AI Act negotiations to the global AI summits. They found 249 cases matching the capture patterns. The dominant ones were not crude lobbying but something subtler: discourse and epistemic influence — controlling the story regulators believe — alongside aggressive legal maneuvering around antitrust, privacy, copyright, and labor. Trade-press coverage reached for the obvious historical parallel, because the researchers did too: this is the tobacco and oil playbook, run again at higher speed.

Inside that mapping, the existential-risk narrative does specific work. Read the mechanism carefully, because it is elegant. If AI is a threat on the order of nuclear war, then only the most capable, best-resourced, most "responsible" actors can be trusted to build it — which is to say, the incumbents. Compliance regimes designed for catastrophic risk are, by construction, expensive: safety evaluations, licensing, reporting infrastructure that a frontier lab can staff and a garage competitor cannot. The sermon about superintelligence and the lobbying position arrive at the same place — a market where the barrier to entry is not capability but a compliance department. Researchers studying the "risk imaginaries" that govern AI policy have documented how the catastrophic framing crowds out attention to present-tense harms — the bias, the labor displacement, the energy burden this desk covers weekly — which happen to be the harms the incumbents are currently causing.

Then there is the revolving door, which needs no theory at all — it is a staffing record. Personnel move fluidly between the major labs and the agencies and offices tasked with overseeing them; the same names recur on both sides of the table. Our Power ledger tracks the specific seats and the specific moves. [VERIFY: name the current documented door-crossings from the Incentive Map dataset before publish — sourced individually, no aggregates]

Did AI do this, or did we?

No model lobbied a regulator. The capture mapping is a catalogue of human choices — executives who discovered that fear is a moat, policy staff who wrote rules their next employer will thank them for, and legislators who found it easier to defer to the loudest experts in the room, who happened to be the vendors. It is worth saying plainly: some researchers voicing catastrophic-risk concerns are sincere, and dismissing all safety work as theater would be its own distortion. The documented problem is narrower — the sincere concern has been harnessed. The people funding the doom conversation are, in the same fiscal year, lobbying against the concrete regulation of their present-tense products. Watch the two hands at once, and the sermon reads differently.

What we are not claiming: that AI carries no risk, or that regulation is unnecessary. The claim is the mapped one: the regulatory framework being built is measurably shaped by the companies it is meant to govern, using a fear of the hypothetical future to protect the economics of the present. The existential risk that is actually documented is to competition, to open research, and to the public's ability to see the industry clearly.

The moat is being poured now, while the audience watches the sky. Our Power lane keeps the ledger on who is pouring it — name by name, filing by filing, door by revolving door.