AI and Your Therapy Practice:
What You Actually Need to Know
AI · ETHICS · PRIVATE PRACTICE NERDERY
This is not legal advice but a way to help provide what’s ethical, what’s legal where you practice, and what to actually do about it.
Legislation table verified monthly - last checked July 2026
The ethical stuff that actually matters
START HERE
Informed consent
If a client’s information is touching an AI tool in any ways, even a transcription tool summarizing notes, they have the right to know about it.
01
Transparency
Consent is “did you tell them.” Transparency is “do they actually understand.” A line buried in a form nobody reads doesn’t count.
02
Confidentiality and privacy
Most consumer AI tools weren’t built with HIPAA in mind. “Which tool has a signed BAA” matters more than “which tool is best.”
03
Misdiagnosis and overreliance
AI is cocky af, even when it’s wrong. The clinical judgement still has to be yours.
04
Surveillance risk
Some state laws exist specifically over where AI-processed data ends up, especially self-harm, substance use or reproductive health info.
05
Abandonment risk
If a client leans on a chatbot between sessions and something goes wrong, “the bot said something harmful” is now a real category of lawsuit.
06
Plagiarism and attribution
If AI drafts your handouts or website copy, be honest about what’s a starting point and what you’re presenting as fully your own thinking.
07
Bias
These tools are trained on existing data, which means existing biases comes along for the ride, especially in anything touching risk assessment.
08
There is no federal law governing AI in therapy. What exists is fifty different answers.
Some states have already enacted laws. Most haven't touched the topic. A few have bills sitting in committee that could change everything by next session. Here's the current spreadsheet below, state by state.
THE FULL BREAKDOWN
Every state, tracked
AI in therapy legislation by state
This table has bill names, effective dates, what’s permitted, what’s prohibited, penalties, and the source links for all 50 states and DC. View here as well. Editable for everyone to contribute.
This table reflects research of the date listed on each row. Laws in this space move fast. Confirm current status with your state licensing board before relying on this for a real decision.
This is not legal advice. If you have a question about your specific situation, that’s a conversation for an attorney or your board.
So, what should you actually do?
THE PRACTICAL PART
- Use AI for admin, not clinical judgment - scheduling, drafting non-clinical content, summarizing your own notes after the fact. Keep diagnostic and treatment decisions yours
- Keep PHI out of consumer AI tools - unless you have a signed BAA with that vendor. Google Workspace with a BAA is legitimate. A free chatbot account is not
- Tell clients if AI touches their care - written consent is mandatory even something as small as an AI-assisted transcription tool. Silence is where the ethical risk lives
- Check your state's status before you build workflows - use the table above first, not after you've already built something around a tool
- Look at your malpractice coverage - ask specifically whether AI-related claims are covered. Some policies haven't caught up to this yet
- Document your reasoning - why you used it, what you reviewed, what you changed. "I didn't think about it" is not the answer you want to give a board
AI isn’t the enemy here, and it isn’t going anywhere. The therapists who come out ahead are the ones who understand the actual rules.
ONE MORE THING
This page will keep growing as the law does. If you’re got a question this didn’t answer, take it to your board, your attorney or your own research. Consider this the on-ramp, not the destination.