The American Nurses Association convened its inaugural AI in Nursing Practice Think Tank on April 22, 2026, and released the consensus report on May 6. The report — developed with nursing leaders across practice, education, research, regulation, industry, and policy — is the first unified professional stance on AI risks and guardrails from organized nursing in the US.

The core rule: AI must support, not replace, professional nursing judgment. Nurses remain the final accountable decision-makers in every clinical scenario. That sounds obvious. The report makes clear it is already being violated — not by malicious design, but by workflow design.

Five Material Risks Already in Clinical Practice

The Think Tank identified five specific risks showing up in real clinical environments right now:

  • Automation bias: Nurses deferring to AI-generated recommendations without applying independent clinical judgment. The algorithmic output becomes the de facto order even when it conflicts with bedside assessment findings.
  • Erosion of professional judgment: Repeated reliance on AI decision support reduces the experiential pattern-recognition that experienced nurses develop over years of clinical practice. Particularly concerning for new graduates entering heavily AI-augmented environments.
  • Unclear license exposure: When an AI system generates a recommendation that a nurse acts on and a patient is harmed, who holds professional liability is genuinely unresolved in most states. State BONs have not issued clear guidance. Nursing malpractice policies have not updated language.
  • Algorithmic bias: AI systems trained on historical patient data inherit the biases in that data — racial, socioeconomic, diagnostic. Nurses who accept AI outputs without critical analysis may unknowingly act on recommendations that disadvantage already-marginalized patients.
  • Alert fatigue amplification: AI-generated alerts layered on top of existing EHR alert systems increase cognitive overload. When nurses silence AI alerts habitually because 90% are false positives, the 10% that matter get ignored.

The Five Near-Term Actions ANA Committed To

The report is not just a risk catalog. ANA committed to five specific near-term actions:

  1. Issue nurse-led guardrails — practice standards for AI interaction that go beyond generic institutional policies
  2. Curate a Nursing AI Playbook — a practical clinical reference for nurses encountering AI tools
  3. Establish AI literacy as a baseline competency — nursing education programs will be expected to address AI literacy, not treat it as optional technology training
  4. Advocate for nursing representation in AI governance structures — bedside nurses must be included in procurement and deployment decisions, not just informatics specialists and administrators
  5. Build an AI incident reporting mechanism — parallel to adverse event reporting, specifically for AI-related clinical near-misses and harms

What This Means at the Bedside

The practical message: you cannot outsource your clinical judgment to an algorithm. If an AI-generated recommendation conflicts with your assessment, your assessment takes precedence. Document both. If your facility's AI workflow is designed to make it structurally difficult to override AI outputs — through extra documentation burden, workflow friction, or supervisory pressure when nurses deviate from algorithmic recommendations — that is a patient safety issue, and ANA is now on record saying so.

The license exposure question is the most uncomfortable one on this list. Most nurses do not know that their malpractice coverage may not explicitly address AI-assisted care scenarios. This is worth a direct conversation with your insurer — ask specifically whether your policy covers adverse outcomes in cases where AI-generated recommendations were part of the clinical workflow. Check the nursing malpractice insurance guide for what to ask when reviewing coverage.

The Nursing Directory Take

This report is meaningful precisely because it comes from organized nursing, not from hospital administrators or health tech vendors who have financial interests in AI adoption. ANA staking out a position that AI cannot override nursing judgment — and committing to enforcement mechanisms through guardrails and governance standards — sets a baseline that nursing leadership at every facility can cite. Use it.

The bigger systemic issue the report flags: most hospitals deploying AI tools in clinical settings are not including bedside nurses in procurement or governance decisions. The people most affected by the tools are not at the table when the tools are selected, configured, and activated. That is not a technical problem. It is a power structure problem, and the ANA is calling it directly.