The Joint Commission Now Has a Rulebook for Hospital AI
要点总结
- 2026 年 6 月 1 日，联合委员会推出了 RUAIH（医疗保健领域人工智能的负责任使用），这是首个专为医疗保健机构打造的此类自愿认证。
- 它认证的是一个组织如何管理人工智能，而不是单个人工智能产品，涵盖五个标准领域。
- 医院、急诊医院和医疗系统均可申请，无论其是否获得联合委员会认证。
- 它与健康人工智能联盟的八项开源治理指南相一致，并且正值FDA人工智能工具和CMS对认证机构更严格的监管等更广泛的举措出台之际。
医院购买人工智能的速度一直超过了其管理能力。这种情况现在已经改变了。
6月1日，负责认证美国大多数医院的联合委员会（Joint Commission）推出了一项名为“医疗保健领域人工智能负责任使用”（Responsible Use of AI in Healthcare，简称RUAIH）的认证。联合委员会称，这是首个专为医疗机构打造的此类认证。该认证为自愿性质，医院无需获得联合委员会的认证即可申请。
有一个细节很容易被忽略，但值得仔细注意：认证并非对单个人工智能产品进行评估，而是认证机构如何管理人工智能——围绕这些工具的政策、监督和流程，而非工具本身。换句话说，它不会告诉你某个脓毒症预测模型是否有效，而是告诉你运行该模型的医院是否有一套切实可行的流程来判断其有效性。
认证实际检查的内容
RUAIH认证围绕五个标准领域构建。联合委员会将其列为：
- 治理
- 有效的数据管理
- 降低风险和偏倚
- 监测、评估和验证安全性能、有效性和负责任使用情况
- 透明度、教育和培训
该认证面向医院、急诊医院和医疗系统开放。联合委员会表示，为了制定该认证，他们召集了20多个拥有医疗保健和技术专长的联盟和团体。联合委员会总裁兼首席执行官乔纳森·B·珀林医学博士（Jonathan B. Perlin, MD, PhD）这样阐述了该认证的目标：“人工智能有潜力带来新的发现，并提高医疗质量、安全性和运营效率。” 在他看来，该认证旨在为各机构提供安全使用人工智能的蓝图。
另一半：共享的行动指南
The certification did not arrive alone. The week before it launched, the Coalition for Health AI (CHAI), the Joint Commission’s partner on this work, released a set of open-source AI governance playbooks for health systems, developed with input from more than 100 healthcare organizations. The two are designed to line up with each other.
The CHAI playbooks break governance into eight elements, including an organizational AI policy, a responsible AI lifecycle, risk and impact assessments, third-party management, and education and training. Brian Anderson, MD, CHAI’s CEO, said the alignment between the playbooks and the certification is meant to reduce confusion and speed up responsible adoption. Aaron Miri, chief digital information officer at Baptist Health, put the need simply: hospitals need a framework to follow for this emerging technology.
The plain-language version: instead of every hospital inventing its own AI rulebook from scratch, there is now a common one to point to, and a certification that checks whether you actually follow it.
Why this lands the same week as an FDA milestone
The governance push is not happening in a vacuum. As 2 Minute Medicine noted in its June 15 AI roundup, the FDA’s Center for Drug Evaluation and Research recently accepted a letter of intent, reported by Reuters on June 3, for an AI-based, “in silico” tool designed to predict drug-induced liver injury. It works by comparing a new drug’s chemical structure against drugs with known safety profiles, and it was admitted to the FDA’s ISTAND pilot program as a first step in a longer qualification process. The stated aim is to catch safety problems earlier and lean less on animal testing.
Read together, the two stories point the same direction. AI is moving from pilot projects into the machinery of care: into how drugs get evaluated and how hospitals make decisions. The infrastructure for governing it is finally being built at the same time, instead of years behind.
One wrinkle in the background
It is worth flagging one thing happening at the same time. On June 12, CMS finalized a rule tightening its oversight of the nine accrediting organizations that survey more than 9,000 Medicare providers, a group that includes the Joint Commission. The rule moves to curb conflicts of interest, including the fee-based consulting that accreditors sell to facilities they later inspect. It does not touch the RUAIH certification, and the consulting limits were held back for another year of public comment. But it is the backdrop this certification launches into: regulators are paying closer attention to how accreditors make money on services around their core accreditation work, and a new paid, voluntary certification is the kind of adjacent product worth watching as that plays out.
Frequently asked questions
What is the Joint Commission’s RUAIH certification?
Responsible Use of AI in Healthcare (RUAIH) is a voluntary certification the Joint Commission launched on June 1, 2026. It is the first certification built exclusively for healthcare organizations, and it certifies governance and processes rather than individual AI products.
What does the certification require?
It is built around five standard areas: governance; effective data management; risk and bias reduction; monitoring, evaluating, and validating safety and performance; and transparency, education, and training.
Do you have to be Joint Commission–accredited to apply?
No. It is open to hospitals, critical access hospitals, and health systems, and an organization does not need to be Joint Commission–accredited to apply.
How does it relate to the Coalition for Health AI (CHAI)?
The week before the certification launched, CHAI released eight open-source AI governance playbooks developed with more than 100 healthcare organizations. The playbooks and the certification are designed to align.
Where we sit, and what it means on the floor
We make scrubs. We do not build clinical AI, and we are not going to pretend to grade a prediction model or tell a hospital how to run its data governance. That is not our lane. But a lot of what gets decided in these frameworks lands on the nurse standing at the bedside, so it is worth reading honestly.
A few honest takes:
- Governance is the right thing to certify first. Most AI failures in hospitals are not exotic; they are a tool quietly drifting out of accuracy with no one assigned to notice. Certifying the “who is watching this” before the “is the tool clever” is the correct order.
- “Voluntary” is the word to watch. A certification only changes practice if enough systems pursue it, or if payers and patients start asking who has it. Worth revisiting in a year to see how many hospitals actually sign up.
- The standard names education and training for a reason. A model is only as safe as the person deciding whether to trust its output at 3 a.m. If the rollout reaches the bedside as a black box, the governance on paper won’t hold.
None of this is ours to certify. But the floor is where it all gets tested, and that is the world we build for. You can see how we think about it at eipnare.com.
Sources: The Joint Commission press release on the Responsible Use of AI in Healthcare certification (June 1, 2026); Coalition for Health AI governance playbooks; 2 Minute Medicine AI Roundup (June 15, 2026); Reuters reporting on the FDA ISTAND liver-injury tool (June 3, 2026); CMS final rule strengthening oversight of accrediting organizations (June 12, 2026).
Edited by Hedy Nie, COO of Eipnare. Connect on LinkedIn.