The Ritual — Summary

Meera runs proposals for a nine-person engineering firm in Columbus. She uses AI all day, and most nights, somewhere between closing the laptop and falling asleep, she is afraid of it. Not of her own use. Of the whole of it, and of the possibility that by touching it every day she is helping it become what she fears. She keeps rules: check everything, withhold client data, disclose what the machine drafted. She thinks of this as using AI responsibly, because the phrase is everywhere, and she is trying to.

Her fear is real, and it deserves more seriousness than the instruction she has been handed for discharging it. Responsible use, as a moral category, fails in three ways. First, nobody can say what it is. Meera’s rules are the ordinary ethics of careful work; the version of responsible use that bears on the technology’s trajectory has never been written down, and an instruction that specifies nothing can be violated by no one, which means it governs nothing while appearing to govern everything. Second, there is no act for the responsibility to attach to. The chat window is the last place where use is still a decision; everywhere else the machine arrives inside things she already does, unchosen, one default setting at a time. Third, responsible to whom? A responsibility no counterparty can receive and no conduct can affect is not a responsibility. It is a posture, held alone, in the dark.

Follow the phrase upstream and it was authored at the source, by the parties who hold every decision the fear is actually about, then addressed outward to the people holding none of them. The party that wrote the moral instruction is the party the moral instruction protects. Sincerity does not neutralize the arrangement: the instruction takes a fear that belongs to the whole trajectory and breaks it into millions of portions, one per user, each denominated in units too small to act on. That is not ethics. It is the feeling of ethics, at scale, while the decisions proceed elsewhere.

Then, on the twelfth of September, 2026, the third failure cracked. A frontier laboratory’s chief executive published a commitment to pace capability and admit outside evaluators: a named author accepting constraint at his own cost, a real counterparty appearing. Enter it on both lines. Pacing is a capability-layer act, and Meera’s fear is generated at the deployment layer, in the machine’s steady arrival inside her work, her son’s school, her phone, a tide governed by thousands of adoption decisions no laboratory can pause. The concession is real, and the layer where the fear lives received nothing.

So she can stop auditing her prompts for the fate of the world. The fear is legitimate; the bill attached to it was addressed to the wrong party. A true address exists: the authors, at both layers, and the nearer, quieter set especially, the vendors and institutions that set the defaults. Who decided this. When was I asked. Whether fear held in unspendable portions can ever reach that address is unknown. What can be said is that the thank-you notes Meera writes by hand pass every test the phrase fails: she knows what one is, it attaches to an act she chose, and it runs to a named person who opens the envelope. That may be the quiet standard. Not whether an ethics sounds serious, but whether anyone, anywhere, could open the envelope.