A customer put the worry plainly in a recent conversation: what stops a learner from opening an AI chatbot mid-course and using it to answer the quiz? For accredited continuing education, assessment integrity is the first thing at risk, and the provider is the one who answers for it.
Learner AI use isn't simply good or bad
The instinct is to treat learner AI use as something to block outright, but it's more complicated than that. In a brainstorming exercise or a practice activity, a learner using AI might be fine, but in a graded assessment that supports a credit claim, the same behavior undermines the result and outcome data. The line depends on the activity, and the provider needs to be able to draw it activity by activity rather than flip one switch for everything.
What ACCME has made clear
In April 2026, ACCME issued a system-wide alert warning that AI-generated content in accredited CE can undermine content validity and put patient safety at risk. Its central message is that the Standards for Integrity and Independence apply regardless of technology, format, or vendor, and that when AI is involved, the accredited provider remains solely responsible for the learner-facing content.
The alert points to concrete expectations, including validating AI outputs before they're deployed and keeping human oversight over what learners see. Responsibility doesn't transfer to the tool or to the company that built it.
The question worth asking a vendor
That changes what a provider should ask a technology partner. "Is your AI compliant?" sounds reassuring, but no vendor can truthfully answer it, because compliance rests with the provider rather than the product. The more useful question is narrower: does the vendor give you the controls, the grounding, and the records to meet your own obligations? That's the standard we're building Cadmium's AI toward.
How the design maps to those obligations
A few of these capabilities are live in Cadmium products today, and one is on the roadmap. Here's how they line up with what an accredited provider is accountable for:
- Control over where AI runs. Turn AI on or off at the account level, so you decide where it belongs and where it stays out entirely. (live today)
- Oversight of behavior. Your team sets the AI's instructions and limits, and every change is versioned with a full history, giving you a defensible record of what the AI was told to do and when. (live today)
- Boundaries by default. The AI only surfaces content a user is already entitled to see, and it doesn't send personal information to the AI models. (live today)
- Validation and monitoring. A separate AI reviewer scores responses for accuracy and safety and compares them across versions, which supports the pre-deployment validation and ongoing monitoring accreditors expect. (live today)
- Grounding in your own approved content. The AI works from your validated materials and accreditation language, so it won't suggest a credit type your organization isn’t accredited to offer.. (on the roadmap)
From risk to signal
The same learner AI interactions a provider worries about are also information. What learners ask, where they get stuck, and how they work through the material all say something about where the gaps are. Handled carefully, that becomes an input to needs analysis, the gap assessment accredited education is supposed to be built on. A course interaction stops being only a risk to manage and starts to inform a stronger, better-documented needs assessment for the next activity.
This is the direction Cadmium is currently designing for, and it's why we talk about AI across the whole accreditation lifecycle rather than only at the guardrail.
The line we won't cross
We won't tell you that any tool makes a provider compliant, because none can. What good design can do is give you the levers to stay in control and the evidence to document your decisions, and that's the bar we're holding ourselves to.
For the fuller picture of the platform underneath these controls, read Why Cadmium Built an AI Platform, Not an AI Feature or speak with an expert here.
.png)
