On: We need ‘right to intervene’ in AI amid growing threat, says Bank of England...

30 September 2026.

The Governor of the Bank of England announces that the authorities must retain a “right to intervene” in the AI industry, lest rogue models take the financial system hostage. One is tempted to applaud. One is also tempted to ask where this right has been for the past decade, while the same industry was invited, subsidised and celebrated into the centre of everything.

The phrasing is instructive. A “right to intervene” is the vocabulary of a man who has already conceded the ground. It presumes the intervention is an exception, a break with a natural order in which the machines arrange themselves and the state merely watches. But there is no natural order here. The architecture of this industry - its concentration, its opacity, its appetite for capital - was built by choices: tax treatment, procurement, the patient tolerance of regulators who mistook scale for sophistication. The Governor asks for a right to intervene in a house whose foundations his own profession helped pour.

Note also what the fear is. Not that the models will fail, but that they will succeed too well - that confidence, that most perishable of commodities, will be entrusted to systems whose expectations no one can audit. A run on a bank is at least a run by persons, who can be shamed, guaranteed, arrested. A run by optimisers answering only to their objective functions offers no such handles. The beauty contest becomes stranger still: the market will price not what others think, but what others think the machines think.

Still, I would rather hear the Governor anxious than serene. Serenity in official quarters is how the disasters of 1929 and 2008 were both prepared. The right to intervene is worth having. The right to have intervened earlier would have been worth more.