AI in Accredited Continuing Education

A question that comes up in almost every conversation with a CME team this year: what stops a learner from opening a chatbot mid-course and using it to answer the assessment? It is a fair worry, and it is also the smaller half of the problem. The larger half is that AI has moved from something your learners might do to something your accreditation now depends on you governing, in writing, with documentation you can hand over on request.

In January 2026 the ACCME published formal guidance on AI in accredited continuing education. In April it followed with a system-wide urgent alert. The alert uses language accrediting bodies do not use casually, including the possibility of suspending accreditation. If you deliver accredited CE and you do not yet have a written AI policy, that is the gap worth closing this quarter.

Key takeaways

• Responsibility does not transfer. ACCME states the accredited provider is solely responsible for learner-facing content, including AI-generated content, and that this does not shift to a vendor, platform, faculty member or partner.
• The Standards still apply. Integrity and Independence apply regardless of technology, format or vendor. AI did not create an exception.
• Seven things must be implemented and documented for any AI-enabled activity, and made available to ACCME on request.
• Accreditation is genuinely at risk. ACCME reserves the right to take immediate action, including suspension, for activities that fail to meet these expectations.
• Your policy needs three audiences. Staff, faculty and learners have different rules, and most draft policies only address the first two.

What did ACCME actually say, and when?

Two documents, and it is worth knowing which is which because they do different jobs.

The first is Guidance on the Responsible Use of Artificial Intelligence (AI) in Accredited Continuing Education, dated 30 January 2026. It is broadly encouraging in tone. It lists uses ACCME actively invites providers to explore, including needs assessments, content drafting, assessment item and distractor generation, evaluation data analysis, and case-building. Then it sets out seven areas of expectation covering independence and bias, disclosure, human oversight, learner privacy, prohibited uses, internal governance and system security.

The second is the Urgent Alert on the Use of AI in Accredited CE, dated 14 April 2026. The tone is different. It warns that poorly designed or inadequately tested AI learning systems may introduce inaccurate, biased or non-evidence-based information into accredited education, compromising learning integrity and potentially threatening patient safety.

The gap between January's encouragement and April's alarm is the interesting part. Something in the intervening ten weeks moved this from guidance to compliance risk, and the alert's focus on learner-facing dynamic AI content suggests what.

Who is responsible when the AI gets it wrong?

You are. The alert is unusually direct about it: if your accredited CE includes AI-facilitated interactions, the accredited provider is solely responsible for the learner-facing content, including any AI-generated content, and that responsibility does not shift to a vendor, platform, faculty member or partner.

Read that against how AI features usually arrive in a CE program. A platform ships an AI summarizer or a chat assistant in a release. Nobody procured it, nobody validated it, and it is now generating learner-facing content inside an accredited activity. Under this guidance that is your content and your exposure, arriving through a product update rather than a decision.

Which makes vendor accountability a contract question rather than a technology question. ACCME expects provider agreements to enable oversight, transparency and control, including access to records of what the AI generated and how learners engaged with it. If your platform cannot produce that log, you cannot meet the expectation no matter how good your internal process is.

What has to be documented for an AI-enabled activity?

The alert lists seven items that providers are expected to implement and document, and to make available to ACCME on request:

• Strict separation from promotion. Learners must be able to engage with accredited content without encountering product-specific marketing.
• Pre-deployment validation. AI outputs validated against defined clinical scenarios before launch, with documented accuracy and completeness.
• Clear guardrails. The scope and limits of AI use defined and aligned to the intended clinical and educational domain.
• Clinical oversight. A named oversight structure with designated clinicians who hold the authority to intervene or discontinue the activity.
• Learner transparency. Disclosure of AI use and its limitations, including known areas of uncertainty.
• Ongoing monitoring. Periodic review, re-validation when systems or prompts change, and prompt correction of errors.
• Vendor accountability. Agreements that give you oversight, transparency, control and access to generation and interaction records.

Notice how many of those are documentation rather than technology. "Designated clinicians with authority to discontinue" is a named person and an escalation path. "Periodic review" is a calendar entry and a record. Most of this is governance work your team can do without buying anything.

So what about learners using AI on your assessments?

Back to the original question, which the guidance answers less directly than providers expect.

The guidance treats learner AI use as a policy matter rather than a technical one. It asks providers to develop policies that differentiate expectations for staff, faculty and learners, and for learners specifically to clarify when AI use is inappropriate due to privacy, copyright or assessment integrity concerns. It also lists auto-producing assessment answers visible to learners among uses that should generally be avoided.

What it does not do is tell you to block chatbots, and that restraint is deliberate. Learner AI use is not uniformly good or bad. A learner using AI to summarize a dense guideline before a case discussion is doing something defensible. The same learner using it to answer the post-test that generates their credit is not. The distinction is the activity type, not the tool.

Practically, that means sorting your activities into three buckets. Formative work where AI assistance is fine and possibly useful. Practice and reflection where it is permitted but should be disclosed. And credit-bearing assessment where it is not acceptable, and where you need something more than an honor system. That is where proctoring, attestation and item design earn their place, and we go deeper on the mechanics in our guide to AI proctoring for continuing education.

What goes in a learner AI-use policy?

Short, specific, and stated before the learner starts rather than buried in terms nobody reads. A workable one covers six things:

1. Scope. Which activities the policy covers, named by type.
2. Permitted use. Where AI assistance is allowed, with an example so the boundary is concrete.
3. Prohibited use. Explicitly including generating responses to credit-bearing assessments.
4. Privacy. No patient information and no PHI into any AI tool, including during case discussions. This is the clause learners breach most often and think about least.
5. Attestation. A statement the learner affirms at assessment, recorded with the attempt.
6. Consequence. What happens to the credit if the policy is breached, decided in advance rather than improvised.

The attestation deserves a note. It will not stop a determined learner, and it is not meant to. Its value is that it converts an ambiguous situation into a documented one, which is precisely what you need when an accreditor asks how you manage assessment integrity. Pair it with the completion and attempt records your platform already keeps and you have evidence rather than assertion.

Frequently asked questions

Does this apply if we only use AI internally, never learner-facing?

Parts of it do. Disclosure, human review of fixed AI-generated materials such as slides and assessments, and screening for bias apply to content development even when learners never touch a tool. The seven-point alert list is aimed at learner-facing dynamic AI.

Do we have to disclose AI use to learners?

When AI was used to generate, modify or analyze educational materials, yes, disclosure is expected. Routine spelling and grammar tools are explicitly excluded.

Our vendor added an AI feature we did not ask for. Are we liable?

Under the alert, the accredited provider is responsible for learner-facing content regardless of its source. Ask your vendor for a written description of what the feature generates and what logs it keeps, and turn it off inside accredited activities until you have that.

Does this apply outside CME?

The ACCME documents govern accredited continuing medical education. Other accreditors set their own expectations, and nursing, pharmacy and allied health bodies are moving in the same direction. If you are jointly accredited, work to the strictest requirement that applies to you.

Where do we start if we have nothing written?

Inventory where AI already touches your activities, including features you did not procure. That inventory usually surprises people, and it tells you which of the seven expectations you are closest to failing.

The bottom line

Accredited continuing education did not get a carve-out from AI, and it did not get a ban either. What it got is an ownership rule: the content is yours, the documentation is yours, and the accreditation risk is yours, whoever built the tool. The providers who will be comfortable at their next reaccreditation are the ones treating this as governance work rather than a technology question, and doing it now rather than during a review.

Most of that work is policy and record-keeping. The part your platform has to carry is the evidence: what the activity presented, what the learner did, what they attested to, and when. If you want to see how that gets captured without adding another manual process, take a look at our healthcare LMS or book a demo and we will walk it against your accreditation profile.