Hermer demands urgent law to control artificial intelligence

The Refrigerator and the Constitution

The official account: the Attorney General of England and Wales, Richard Hermer, warns that artificial intelligence is less regulated than a refrigerator, that the risk to humanity is without precedent, and that the remedy lies in urgent public debate and robust international law. The machinery is something else. The comparison he chose shows, more plainly than he may have intended, where governing power actually sits. The gap between the two is not hypocrisy. It is how the system works, and understanding it is more useful than lamenting it.

Begin with the refrigerator, because Lord Hermer was right to choose it, though perhaps not for his reason. Nobody ever held a public debate about refrigerators. No convention was signed at Geneva on the proper temperature of a salad drawer. The refrigerator is governed through the efficient parts of the state alone: a technical standards committee nobody has heard of, an electrical safety mark, a product liability rule that makes the manufacturer pay when the compressor catches fire, an insurer who will not cover a factory that ignores the rules, and a retailer who will not stock an appliance without the right sticker on the door. The refrigerator is safe because a great many dull people have a commercial reason to make it so. Its regulation has no dignity at all, and that is why it works.

Now consider what the Attorney General proposes for artificial intelligence: public debate and international law. These are the most dignified instruments a government possesses. They are the coronation and the State Opening of regulatory life. They exist to command reverence, signal seriousness, and confer legitimacy, and they do these things well. What they do not do on their own is make anything happen on a Tuesday afternoon in a data centre.

The strongest objection deserves a fair hearing, and it is a serious one. A refrigerator does not cross borders under its own power or improve itself overnight. If the danger is that humanity may lose control of its future, a domestic safety mark is a pea-shooter against a tide. The risk is transnational, so the law must be. Only states acting together can bind firms that can relocate their servers faster than a ministry can draft a consultation paper. On this view, international law is not ceremony. It is the only instrument big enough for the problem.

The objection is right about the scale and wrong about the mechanism. The international agreements that have actually constrained dangerous technologies did not work because they were robust in their language. They worked because each was attached to something unglamorous: inspectors, export licences, shipments of controlled material that could be counted at a port. Where an agreement had no such attachment, it produced fine communiqués and very little else. The questions that matter about artificial intelligence are operational ones. Who supplies the chips, who rents out the computing power, who insures the deployment, and who is liable when the system harms someone? None of these is settled by a debate, and all of them are settled, at present, by contract. The terms of service of a handful of companies are the working constitution of the field. Their published principles of responsibility are its dignified part.

Here the true convention appears. Artificial intelligence is not unregulated. It is regulated privately, by the firms, their cloud providers and their lawyers, with the state as an occasional spectator. A treaty layered on top of this arrangement would change its appearance far more than its operation. The firms would sign up, with every sign of sincerity, to principles they already publish.

None of this makes the Attorney General’s speech useless. Dignified institutions build the confidence that efficient ones later spend. Ministers cannot impose liability on a fashionable industry until the public has been persuaded that it is dangerous, and a law officer comparing it unfavourably to a kitchen appliance is exactly the sort of remark that does the persuading. The ceremony prepares the ground. The danger lies in mistaking it for the work. A government can hold the debate, sign the convention, and feel it has acted, while the real decisions continue in procurement offices and indemnity clauses it never thought to read.

So the refrigerator teaches the opposite lesson from the one intended. It is not a reproach for how little attention artificial intelligence receives. It is a model of how much can be achieved with no attention at all, provided someone dull is made liable. Somewhere in a showroom, a shopper opens a fridge door, glances at the energy label, and closes it again, perfectly confident. That shopper has never read a treaty. They have never needed to.