On: Bernie Sanders speaks to the BBC about proposal to ban AI superintelligence

Eleventh of September, 2026.

An old man in Congress proposes to forbid the building of a mind greater than ours, and to have the government buy half of the minds already being built. I read this twice, and admired it both times, for different reasons.

The first reason is that the proposal is honest. For a century the statesmen have pretended to govern the new forces while merely standing near them, the way a stationmaster stands near an express train he cannot slow. Here at last is a man who says plainly: this force should not exist, or if it must exist, the public should hold the reins and the dividends. Whether one agrees is beside the point; it is the first sentence addressed to the force itself rather than to its by-products.

The second reason is darker. The proposal assumes that a legislature can prohibit a capability the way it prohibits a distillery or a lotteries act. But capability does not wait upon statute. It arrives at whatever velocity the physics and the capital allow, and the law arrives later, on the older schedule, with its leather satchel and its paper forms. A ban on superintelligence passed in 2026 will be enforced against 2026, and will meet 2032 as a customs officer meets a jet aircraft - with great dignity and no jurisdiction.

Still, I confess a certain tenderness toward the effort. My own class was trained to govern a republic of railroads and treaties, and we governed it beautifully, right up to the moment it stopped existing. Senator Sanders was trained for trust-busting, and he is applying trust-busting to a thing that is not a trust. The categories are wrong. But wrong categories, energetically applied, are the only instrument the species has ever had. The Virgin organised an age she did not understand either.

Perhaps the halting is the point. Perhaps a law that fails slowly is still a brake.