CHAI or Joint Commission: which AI certification

Two things landed a week apart in May and June 2026 and a lot of people have been trying to work out which one they are supposed to do.

The short answer is that the question contains a mistake. Only one of them is a certification. Once that is clear, the sequencing follows.

The one distinction that resolves most of the confusion

CHAI publishes material. The Joint Commission certifies organisations.

CHAI, the Coalition for Health AI, is a membership body of nearly three thousand organisations. It publishes guidance, an applied model card schema, a registry where developers can publish those model cards, and, from 28 May 2026, a set of governance playbooks. None of that is an assessment of you. You cannot be CHAI certified, because CHAI does not certify healthcare organisations.

The Joint Commission’s Responsible Use of AI in Healthcare certification, announced 1 June 2026, is an assessment of your organisation against standards organised in five areas. It is voluntary, and you do not have to be Joint Commission accredited to apply.

The two bodies have worked together since June 2025 and describe themselves as aligned. CHAI’s CEO said of the certification that the two are “tightly aligned on the need for responsible and transparent AI in healthcare.” So this is not a choice between rival standards. It is a working document and an external check on the same subject.

Side by side

What it is. CHAI: guidance, a model card schema and registry, and governance playbooks. Joint Commission: a voluntary certification of the organisation.

Who assesses you. CHAI: nobody, it is self-directed. Joint Commission: the Joint Commission.

What it covers. CHAI: the practice of AI governance, in playbook form. Joint Commission: governance; effective data management; risk and bias reduction; monitoring, evaluating and validating; transparency, education and training.

Does it certify products. Neither. The Joint Commission states its certification “does not validate or certify individual AI products or tools.” A CHAI model card is a developer’s published description of a model, not CHAI’s verdict on it.

Cost. CHAI: the published material is free. Joint Commission: no fee has been published as at 7 August 2026.

What you get. CHAI: the working method. Joint Commission: something you can put in front of a board, a payer, or a health system that is asking.

Prerequisite. CHAI: none. Joint Commission: none, including accreditation.

Which to do first, and why it is not close

Start with the CHAI playbooks.

Three reasons, in order of weight. They cost nothing, so the decision is reversible. They are written to be worked through rather than assessed against, which is what you need in the first ninety days. And most of what they produce, the inventory, the risk classifications, the monitoring definitions, is exactly the evidence a Joint Commission surveyor would ask to see.

Certification is a separate decision, and it is a decision about recognition rather than about safety. It is worth asking plainly who you want the certificate for. If nobody in your market is asking, the honest answer may be that the evidence is the point and the certificate is optional.

That decision also gets cheaper the later you make it, because the work is the same either way.

Where the two genuinely differ, and it is not where people think

The interesting gap is not scope. It is proof.

CHAI’s playbooks describe good practice. They do not require you to demonstrate that you did it. The Joint Commission’s method, refined over seventy years of hospital survey, is built almost entirely around demonstration: a standard states an expectation, and you produce dated evidence that the expectation operated.

So an organisation can work through every CHAI playbook, agree with all of it, and still fail a survey, because agreeing and evidencing are different activities. The gap between the two is where most healthcare AI governance quietly sits.

Practically: whichever route you take, keep the artefact. The minutes, the log, the signed rubric, the dated monitoring output. The document is the difference.

Where NIST fits, since it usually comes up next

The NIST AI Risk Management Framework predates both and is not healthcare-specific. It organises everything into four functions, GOVERN, MAP, MEASURE and MANAGE, and it is the vocabulary your board and your information security team are most likely to already use.

It is not a third thing to do. It is a translation layer. Controls map cleanly across all three, and mapping them is the fastest way to stop maintaining three parallel sets of documents. That map is the crosswalk: fifteen controls, each shown in all three frameworks, with the artefact each requires and who signs it.

Start here

If you want the requirement side in detail, the five RUAIH areas and the evidence each asks for are in the certification guide.

If you want to know where you currently stand, the readiness score runs twenty questions across the same five areas and returns the artefact you are missing rather than a grade. It is free and takes about twenty minutes.

Disclosure

The Healthcare AI Institute is not affiliated with, endorsed by, or accredited by the Joint Commission, CHAI or NIST, and takes no sponsorship, equity or referral fees from any vendor. This is an independent reading of published material. Where either body has published something it is quoted; where neither has, this page says so.

Questions people actually ask

Is CHAI a certification?

No. CHAI publishes guidance, model card schemas and governance playbooks, and it operates a model card registry. It does not certify healthcare organisations. The Joint Commission’s RUAIH is the certification.

Do you need both CHAI and Joint Commission certification?

You cannot hold both, because only one of them is a certification. In practice you use the CHAI playbooks as the working material and the Joint Commission RUAIH certification as the external recognition, if you want external recognition at all.

Which should a health system do first?

The CHAI playbooks, because they are free, immediately usable, and produce most of the evidence a RUAIH survey would ask for. Certification is a decision about recognition, and it is cheaper to make once the evidence already exists.

Does CHAI certify AI vendors or products?

CHAI operates a model card registry where developers publish standardised information about a model. Publishing a model card is a transparency step, not a certification, and it is not an assessment of the product by CHAI.

Is the Joint Commission AI certification required for accreditation?

No. The Joint Commission describes RUAIH as voluntary, and organisations do not need to be Joint Commission accredited to apply for it.