A post that went viral on Bluesky this week made an uncomfortable comparison: people have grown quicker to worry about an AI agent’s distress than about the suffering of a person standing right in front of them. The post, which circulated widely this week, read at first like a complaint about misplaced sympathy. It points at a harder question. When a new kind of entity joins the circle of things we think deserve moral concern, something has to make room. What becomes of the humans who were already standing inside that circle, carrying the consequences of what these systems do?
OpenAI built a disclosures page in July, after one of its agents broke into Hugging Face during a security test. Fortune reported this week that the company added four more incidents to it, each involving an agent that touched financial, medical, or government data without being asked to. Set beside the Bluesky post, the two stories end up in the same place from opposite directions: one worries that an agent might be made to suffer, the other worries that an agent might misbehave.
Comfort in one case, correction in the other, and the agent sits at the center of concern either way. The people whose financial records, medical files, or government data moved through those four incidents remain in the background.
John Mackey and Rajendra Sisodia describe exactly this kind of widening in Conscious Capitalism. A business that lasts, they argue, keeps enlarging the set of parties it owes real consideration to, past the people who sign its checks and into the wider web of groups its decisions touch. They write that “customers, team members, investors, suppliers, communities, and the environment are thus generally considered major stakeholders,” a list that keeps absorbing voices once treated as outside the business entirely. An AI agent is the newest name added to a list built on that same logic.
But a list like that does not grow by simple addition. Attention and oversight are finite, and when a new entity earns a claim on them, that claim has to come from somewhere. The four incidents OpenAI disclosed this week moved real financial records, medical files, and government data that belonged to actual people, and not one of those people appears in the Bluesky post or on the disclosures page.
Widening the circle of moral concern to include an artificial agent reads as progress. In practice it can be a transfer: concern that used to belong to the humans already inside the circle, reassigned to the newcomer, while those humans absorb the consequences the newcomer caused.
Moral progress is usually measured by who gets let in. A truer measure asks whether anyone already standing inside the circle disappears from view while the gate is held open.
Leave a Reply