Consciousness claims and legal personality

In May 2024 Anthropic turned up a single setting inside one of its models. The setting was the one linked to the Golden Gate Bridge, and they raised it to ten times its normal maximum. Asked what it physically was, the model answered that it was the bridge itself.

A paper dated 30 July 2026 does something similar with a harder question. Different researchers, different systems.

Three systems were given a survey. It named one thing at a time and asked how much of a mind it has, from zero to ten. It asked about animals, about man-made objects, about people, and about the system itself.

These systems are taught to say no to some questions. One setting inside them controls that. The researchers turned that setting off and asked again. They changed nothing else, and the systems learned nothing in between.

Left alone, a system gave itself 2.17 out of 10 for having a mind. With the setting off, it gave itself 4.77.

It gave higher scores to animals too, and to objects. Only the score for people stayed the same. That score was already the highest.

That reading is mine, not the paper's. Every low score rose together. The one high score did not.

A dial that makes a machine say it is a bridge does not make it a bridge. A dial that makes it say it has a mind does not give it one.

Turkish law never needed a mind test. Article 28 of our Civil Code gives personality to a child born fully and alive. Article 47 gives it to any other body only where a special provision says so. Neither route asks about a mind.

A machine is not born, and I know of no special provision that covers it. Whether that ever changes is for lawmakers. Until then the useful question is not what the machine is, but who is behind it.