Maine Restricts AI in Therapy, But Loopholes May Undermine Intent

Maine has enacted legislation restricting how therapists can deploy artificial intelligence in mental health treatment, joining a growing number of U.S. states attempting to regulate AI's role in clinical settings. While the law aims to prevent AI from replacing human judgment in therapy, its language contains ambiguities that could complicate enforcement and potentially allow workarounds.

What the law prohibits

According to reporting by Forbes contributor Lance Eliot, Maine's statute explicitly bars AI systems from making independent therapeutic decisions, directly engaging clients in therapeutic communication, or generating treatment plans without professional review. Licensed mental health professionals may still use AI for administrative tasks or supplementary support, provided they maintain full responsibility for all clinical decisions.

The restrictions reflect widespread concern about AI systems operating autonomously in sensitive mental health contexts, where misdiagnosis or inappropriate guidance could cause significant harm.

The internet-based AI loophole

A notable weakness in the legislation is its specific reference to "Internet-based artificial intelligence." This phrasing potentially exempts standalone AI tools that operate without internet connectivity—a distinction that may not align with lawmakers' intentions. As AI applications increasingly function offline or through hybrid architectures, this language could create an unintended carve-out for certain technologies.

The law also employs a broad definition of "therapeutic communication" without clear boundaries, leaving therapists and AI developers uncertain about which interactions fall under the prohibition. This vagueness could lead to inconsistent interpretation by prosecutors and compliance challenges for practitioners trying to follow the rules.

Pattern across state legislation

Eliot notes that Maine's approach mirrors AI mental health laws passed in other states, reflecting a nationwide effort to establish guardrails around therapeutic AI. However, the rush to regulate rapidly evolving technology has produced legislation that often contains inadvertent loopholes or insufficiently precise language.

These drafting issues create risk on both sides: AI developers and therapists might unknowingly violate restrictions based on one reasonable interpretation, while others could exploit ambiguities to circumvent the law's protective intent.

Why it matters

As generative AI capabilities expand, mental health applications represent one of the highest-stakes domains for regulation. Poorly crafted legislation doesn't just fail to protect patients—it creates legal uncertainty that can stifle legitimate innovation while leaving dangerous applications unregulated. Maine's experience illustrates the difficulty of writing durable rules for technology that evolves faster than the legislative process. States will need to develop more precise regulatory frameworks and update them regularly, or risk laws that are either too restrictive for beneficial uses or too porous to prevent harm.

The details of Maine's AI mental health law were first reported by Lance Eliot in Forbes.