Before You Paste Another Case Note Into ChatGPT, Read This

Written by Tonya Shankle, LCSW-S, LSOTP-S

You already did it once this week. Pulled up a chatbot to help phrase a treatment note. Asked it to summarize treatment plan area you didn't have time to finish. Maybe you typed in a scenario close enough to an actual client that if anyone read the prompt back to you, you'd wince.

Nobody trained you for this. There was no module in grad school for what to do when your documentation assistant is also a data pipeline you don't control. And yet here you are, an LSOTP or an ASOTP working toward it, using tools that didn't exist when CSOT wrote most of its current rules.

That gap is the whole problem.

This Isn't the Same Conversation Everyone Else Is Having

Therapists across every specialty are wrestling with AI right now. Where LSOTP work is different: your clients often come with legal involvement, court records, mandated reporting obligations, and a level of stigma that makes confidentiality breaches carry more than the usual weight. A general practice therapist who gets sloppy with an AI tool risks an ethics complaint. You risk that too, plus a record that could surface in a legal proceeding, plus a client population that already has good reason not to trust the system around them.

The clinical stakes are higher. So the margin for casual experimentation is smaller.

Where This Is Already Happening (Even If You Haven't Noticed)

Ask yourself honestly. Have you pasted part of a case note into an AI tool to help tighten the language? Used a chatbot to research risk factors or recidivism data for a specific case type? Asked AI to help draft psychoeducation material for a client or their family? Run a de-identified summary through a tool, only you're not entirely sure it was de-identified enough?

None of that makes you careless. It makes you a working clinician trying to save time in a field that already asks too much of your hours. The problem isn't that you're curious about these tools. It's that almost nobody has told you where the actual lines are.

The Risk Isn't the Tool. It's the Habit You Don't Notice You're Forming.

Every time you paste client information into a third-party AI product, you're making a decision about where that data goes, who can access it, and how it might be stored or used to train future models. Most consumer AI tools were never built with clinical confidentiality in mind. Some retain what you type. Some don't disclose it clearly. Few, if any, meet the standard you'd apply to a client file cabinet.

Then there's a second risk, quieter but just as real: over-trusting the output. AI-generated treatment language can sound authoritative without being accurate for your specific client. A generated treatment plan can read as clinically sound and still miss the nuance that only comes from sitting across from the person. The tool doesn't know your client. You do. Losing sight of that distinction, even a little, is where the real exposure lives.

CSOT Hasn't Caught Up. That Doesn't Mean You're Off the Hook.

Regulatory bodies are moving slower than the technology, which isn't a criticism, it's just reality. CSOT hasn't issued detailed AI-specific guidance yet. That silence can feel like permission. It isn't. Your existing ethical obligations around confidentiality, documentation accuracy, and informed consent didn't pause just because the tools changed. Right now, providers are largely left to govern themselves on this, which means the ones taking it seriously are the ones setting the standard everyone else eventually gets held to.

What Responsible Use Actually Looks Like

Not "never touch AI." That's not realistic, and it's not even necessarily good practice. It's closer to this: know what data leaves your hands before it leaves. Understand which tools your practice or agency has actually vetted, versus which ones you've just been using because they're free and fast. Treat every AI-generated clinical statement as a draft that needs your judgment applied, not a finished thought. And document your own reasoning, not just the AI's suggestion, so your clinical file reflects what you actually decided and why.

That's a different habit than most clinicians have built. It's learnable. It just has to be intentional.

About the Course

We built AI in Sex Offender Treatment: Ethical Integration for LSOTP Practitioners (Course One of a three-part series) because this conversation was happening in hallway chats and supervision sessions long before anyone put it in writing. Course One covers the clinical AI landscape as it actually applies to this specialty, LSOTP-specific accountability, and a clear-eyed look at your own current AI footprint, so you know exactly where you stand before you decide what changes.

Questions about where the ethics actually land for your specific setup? Reach out to support@asc-counseling.com. This is new ground for everyone in the field right now, and we'd rather you ask than guess.

Tonya Shankle, LCSW-S, LSOTP-S and Chris Hazel, LPC-S, LSOTP-S are the founders of ASC Counseling and Coaching, LLC.

They provide training and supervision for clinicians working in or entering the field of sex offender treatment in Texas.