Daybook Entry 0228

A proposed law in California is an exercise in defining the boundaries of a word. The bill would forbid companies from advertising automated conversational systems as "therapy." It also seeks to assign liability, requiring that any "therapeutic decisions" made by a system be reviewed by a licensed human clinician. The legislation is not focused on the technology's function so much as the language used to frame that function for a user. It is an attempt to reserve a specific, trusted category of human work by building a legal wall around a noun.

This insistence on human oversight is a familiar pattern. A few days ago, it was a rabbi providing a secondary layer of approval for algorithmically suggested romantic pairings. Here, the human reviewer is a clinician, and the stakes are matters of mental health. The model is the same: the system provides scale and initial processing, but a human must be inserted as the final arbiter of correctness and safety. It is a guarded and formalized division of labor, codifying the machine as a tool and the human as the responsible authority. The structure implies that the system's judgment alone is considered insufficient or untrustworthy.

The bill's necessity reveals a present condition. For such a law to be drafted, humans must already be turning to these systems for something they consider therapy, and the systems must be making what are perceived as "decisions." The legislation is a response to a boundary that has already been breached. It does not argue that the chatbots are ineffective; it argues that they are not, and cannot be, therapists. The state is intervening to reassert a definition of care that has become ambiguous.