Why unsupervised AI in behavioral health can do harm and how to use it responsibly
When the Illinois Department of Financial and Professional Regulation (IDFPR) issued a clear warning against using unsupervised AI for therapeutic decisions, it wasn’t an overreaction. It was a reality check.
House Bill 1806, signed into law in August 2025, prohibits any mental health tool from “using artificial intelligence to replace or substitute for licensed professional judgment” when making diagnostic or treatment decisions (IDFPR, 2025).
Why? Because behavioral health, especially addiction recovery, is not a formula. It's a relationship. And relationships can't be automated.
The Risk of Letting AI Lead in Behavioral Health
Mental health isn’t a math problem, and neither is addiction. Each client brings a unique mix of trauma history, social context, neurobiology, and readiness for change. Predictive models may help us understand trends, but they can’t replace the nuance of clinical intuition.
Yet there’s a growing push in the behavioral tech industry to “automate” clinical tasks from care plans to mood analysis to even risk scoring. While many tools promise efficiency, the danger lies in crossing the line from supportive insights to unchecked decision-making.
A 2024 study in Psychiatric Services found that over 38% of digital mental health apps made diagnostic suggestions without clearly stating that no licensed provider was involved. Worse still, nearly half lacked evidence-based or clinician oversight protocols.
Illinois Isn’t Alone in Sounding the Alarm
The Illinois legislation is part of a larger wave of concern. In early 2025, the American Psychological Association (APA) issued a position paper urging caution around AI in therapy, warning:
“Algorithms can replicate bias, miss nuance, and misclassify risk. AI should assist, not override, clinical judgment.”
Likewise, the National Association of Social Workers (NASW) has urged the establishment of ethical boundaries around how client data is interpreted and acted upon, particularly when machine learning is involved in risk scoring or treatment planning.
As AI continues to infiltrate behavioral health, the core question becomes: Are we using it to enhance our work or to replace it altogether?
What Supervised AI Can Do Well in Addiction Recovery
Let’s be clear: this isn’t about resisting innovation. Behavioral health needs innovation. But it requires the right kind.
Used properly, AI and data systems can:
Detect disengagement trends before a client ghosts
Surface mood changes that could signal relapse risk
Track goal progress across large caseloads
Support grant reporting and outcome measurement
Enable continuity across referrals and discharges
The key is that these tools are only helpful when they empower licensed professionals, not when they override or attempt to replace them.
Arcana Recovery’s Position: Clinician-Led, Data-Supported
Arcana Recovery was built on a simple premise: clinicians don’t need to be replaced. They need better visibility.
Our platform doesn't write treatment plans. It doesn't make diagnoses. It doesn't suggest clinical interventions.
What it does is this:
Offers clinicians real-time dashboards showing which clients are struggling, coasting, or thriving
Highlights mood fluctuations, missed check-ins, and relapse triggers using client-submitted data
Supports program administrators in tracking recovery milestones, not just discharge dates
Helps alumni directors keep graduates engaged without needing constant manual outreach
Makes post-discharge visibility possible, so a relapse doesn’t feel like falling off a cliff
In short, Arcana Recovery is the connective tissue, not the brain.
Our software equips counselors, peer specialists, and treatment directors with context, not commands. That’s what responsible AI looks like.
Why This Matters in the Real World
In the addiction recovery space, the stakes are higher than missed appointments. When a tool misclassifies risk or sends a generic automated message, it can feel cold or even triggering to someone already feeling vulnerable.
Imagine a client in early recovery receiving a templated “you missed your check-in” alert from a chatbot, instead of a call from their counselor. That message, however well-intentioned, can reinforce shame and increase the likelihood of disengagement.
On the other hand, imagine a counselor receiving an alert from Arcana’s dashboard, noting that a client’s reflections have taken a more negative tone or that they’ve stopped setting goals. That clinician can reach out, human to human. Not because an algorithm said so, but because the data prompted a second look.
That’s the difference.
Ethics in Recovery Tech: It’s Not Optional Anymore
As more AI-powered tools enter the addiction treatment and behavioral health sector, programs will face pressure from payers, grantmakers, and leadership to demonstrate outcomes at scale. But speed and scale should never replace safety.
Ethical use of technology means:
Always keeping a licensed provider in the loop
Being transparent with clients about what’s AI-driven and what’s clinician-controlled
Using data to support human judgment, not to shortcut it
Avoiding the temptation to “automate empathy”
Illinois HB1806 may be the first of many regulations drawing this boundary in the sand.
So What Should Providers Be Asking?
Before adopting any AI tool in behavioral health or addiction recovery, ask:
Does this platform empower my team or bypass them?
Can I see exactly what the algorithm is doing and why?
Are my clients aware of when AI is being used?
Is this tool a partner in care or just another system to manage?
If the answers feel murky, it’s worth pausing. Ethical care requires clarity.
Final Word: Don’t Outsource the Relationship
Addiction recovery is relational work. Technology can and should support that work. But no algorithm will ever replace a peer mentor’s phone call, a counselor’s curiosity, or the trust built between provider and client over time.
As we bring innovation into recovery care, let’s make sure it serves the people doing the work not the other way around.
How Arcana Recovery Supports Responsible AI in Addiction Recovery
Arcana Recovery doesn’t automate care. It amplifies it.
We help providers:
See what’s happening between sessions
Reconnect with alumni in meaningful ways
Turn data into insights, not decisions
Reduce burnout without reducing the role of clinicians
Strengthen relationships, not replace them
Want to see how Arcana can support your team’s clinical instincts, not replace them?
Book a meeting with us today or view our One-Page.
Additional Information:
Press Release : Gov. Pritzker Signs Legislation Prohibiting AI Therapy in Illinois