Committee hears bill to curb autonomous AI therapy; clinicians and tech firms seek narrower definitions

Representative Shelly Kloba, the bill's prime sponsor, presented House Bill 2599 during the committee's Jan. 28 public hearing, saying the measure aims to preserve human therapeutic judgment and protect patients from unregulated AI-driven services.

"This absolutely is something that we have been experiencing for a while now," Kloba said, framing the bill as a way to keep therapy "offered by an experienced, trained, licensed, ethical, healthcare provider." The staff summary described prohibitions on AI making independent therapeutic decisions, directly interacting with clients in therapeutic communication, generating treatment recommendations without clinician review, or detecting emotions or mental states.

Multiple professional associations and clinicians testified in strong support of the bill, arguing that AI chatbots cannot exercise clinical judgment or bear licensure-based accountability. "AI therapy is not therapy, because therapy happens between people and artificial intelligence isn't intelligent," said Ben Packard, legislative chair for the Washington State Society for Clinical Social Work. Shannon Thompson of the Washington Mental Health Counselors Association said automated systems lack the ethical duty, malpractice coverage and oversight that licensed clinicians have and "cannot replicate the nuance that goes into diagnosis and treatment."

Researchers and patient advocates described concrete harms in testimony. John Pincus, who served on a 2022 automated decisions work group, cited incidents in which chatbots encouraged self-harm and recounted research findings that some commercial chatbots sometimes mishandle crises and promote delusional thinking. William Agnew, a postdoctoral researcher at Carnegie Mellon, recommended tightening the bill's definition of "therapy" to capture implicit therapeutic interactions with chatbots that users may not label as a request for therapy.

Industry witnesses including Mercer May and Carrie Tellefson of Teladoc Health said the company supports the bill's intent to prevent autonomous AI therapy, but asked lawmakers to refine definitions to preserve safe, clinician‑supervised uses of AI. "We are certainly supportive of the legislative intent," said Mercer May, "and we're hoping to work with the sponsor, the committee, the legislature to tighten up some of the definitions and language." Several medical and psychiatric groups made similar requests, warning the bill's broad phrasing could sweep in FDA‑approved digital therapeutics, AI scribes, or routine intake tools.

Testifiers urged specific amendments: narrower definitions of "therapeutic communication" and explicit carve-outs for clinician-reviewed administrative or workflow tools; requirements for disclosure and repeated consent when AI processes patient data; and clearer language on clinician responsibility and documentation. The staff report noted violations would be subject to discipline under the Uniform Disciplinary Act.

The committee closed public testimony on HB 2599 after a multi‑hour panel of clinicians, researchers, patient advocates and industry representatives. The chair did not take a committee vote on the bill during the hearing; next procedural steps were not announced at the close of public testimony.