Europe debates AI regulation to manage global risks
The principle operating here, stated plainly, is: regulate the artificial intelligence that others will build, deploy, and profit from, so that its risks are borne collectively while its exemptions are reserved for the regulator. Let us ask whether this principle, universalised, produces coherence or contradiction.
Europe’s lawmakers, gathered in the ordinary business of law-making on this 25th of August 2026, are not merely drafting technical schedules and risk categories. They are, whether they know it or not, testing a maxim before the world has finished writing it down. The maxim is this: that a legislature may impose upon private actors constraints on the use of a technology which it declines to impose upon itself, on the ground that its own use of that technology serves ends too important to submit to the general rule. This is not a hypothetical. It is the structure of every draft that carves out a national-security exception, every clause that permits a state ministry to deploy a system that a hospital or a hiring firm could not lawfully touch.
Universalise it. Suppose every legislature in the world adopted this maxim: bind the citizen, exempt the sovereign, wherever the sovereign judges the exemption necessary. The result is not a world with less risk from artificial intelligence but a world with the same risk relocated to the one actor least accountable for bearing it. A rule that a border-control algorithm, a predictive-policing system, or a state surveillance tool may operate outside the very safeguards devised for the marketplace is not a rule at all. It is a maxim that treats “necessity” as a private password that unlocks what public law forbids everyone else.
Consider the clerk who sits in the committee room translating the day’s testimony into schedules of permitted and forbidden uses. She adds an exemption for a facial-recognition system used at a border crossing, not because the danger to the traveller passing through it is smaller, but because the state’s interest in using it is larger than the state’s interest in restraining itself. Her pen does not distinguish; her exemption clause does. That is the whole difficulty condensed into a single administrative gesture.
The humanity formula demands more than caution. It demands that the traveller at that crossing, and the applicant whose resume is filtered by an algorithm, and the citizen whose face is matched against a database, be treated as ends in themselves and not merely as data through which the state pursues its own governance efficiently. A regulatory framework that manages risk for the market while quietly exempting the state’s own instruments of power treats the governed as means twice over - once as subjects of the algorithm, and once as subjects of a law that was never meant to bind their own government.
None of this argues against regulation; it argues for a regulation honest enough to bind its author. The lawmakers of Europe are attempting something genuinely difficult: a framework meant to manage risk and secure benefit not for one nation but, given how interconnected these systems are, for economies and societies well beyond their own borders. That ambition deserves respect. But ambition without symmetry produces a structure that looks like law and functions like privilege.
The duty that follows is exacting and simple: any restriction fit to bind the firm must first be tested against the ministry, and any exemption too dangerous to grant a corporation is too dangerous to grant a state. Until that symmetry is written into the text, the clerk’s pen will keep moving, and the exemption clause will remain the one part of the law that never has to justify itself to the very citizens it was written to protect.