Association of periOperative Registered Nurses

See attached file(s)

Considerations for the Regulation of Generative AI-Enabled Medical Devices
Association of periOperative Registered Nurses (AORN)
The Association of periOperative Registered Nurses (AORN) appreciates the opportunity to
comment on the regulation of generative artificial intelligence (GenAI)-enabled medical
devices from the perspective of perioperative nursing practice, patient safety, and
responsible AI integration.

AORN’s Guideline for Integration of Artificial Intelligence and Position Statement on
Artificial Intelligence (AI) in Perioperative Nursing emphasize governance, transparency,
accountability, human oversight, privacy and security, equity, education, and ongoing
evaluation across the technology lifecycle. AORN maintains that AI should support, not
replace, perioperative RN expertise, judgment, empathy, and accountability (AORN, 2026a,
2026b).

Overarching Principles
AORN’s guidance supports AI integration that complements professional judgment; uses
lifecycle governance; preserves human oversight and critical review; promotes
transparency, explainability, accountability, privacy, security, equity, and trustworthiness;
includes role-based education and competency verification; and monitors for safety
concerns, unintended consequences, and performance changes (AORN, 2026a, 2026b).

Responses to Selected Discussion Questions
Question 1: Risk Framework Dimensions
AORN Position
AORN supports FDA's proposed two-axis framework that considers device activity and the
consequences of reliance on an incorrect output. AORN believes these dimensions provide
a useful foundation for assessing risk associated with GenAI-enabled medical devices.

Rationale
AORN guidance emphasizes that risk depends not only on an AI output, but also on
whether clinicians can understand, evaluate, monitor, and respond to that output through
transparent, accountable, and governed processes (AORN, 2026a, 2026b).

Response: “Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper
and Request for Feedback,” Docket No. FDA-2026-N-7874
Recommendations for FDA Consideration
• The degree of human oversight required before clinical action is taken.
• The ability of end users to critically review, question, override, or reverse AI-generated
recommendations.
• The transparency, explainability, and interpretability of outputs.
• The extent to which clinicians can identify errors, uncertainty, or limitations.
• The impact of the technology on workflow, communication, and team decision-making.
• The organization's ability to monitor performance and manage risk throughout the
technology lifecycle.
• The clarity of accountability among developers, vendors, health care organizations, and
end users.

Illustrative Example
A system that allows clinicians to understand and evaluate a recommendation may
present different practical risk than a system that produces highly directive, nontransparent recommendations. The ability to identify and address potential errors before
they affect care should factor into risk assessment (AORN, 2026a, 2026b).

Question 3: Patient-Facing Versus Health Care Professional-Facing
Functions
AORN Position
AORN recognizes the potential for AI-enabled technologies to improve patient
engagement, communication, and access to health information. Patient-facing systems
may also require safeguards that account for whether an intended user can understand the
output, its limitations, and when involvement of a health care professional is needed.

Rationale
AORN guidance supports transparency, privacy, trustworthiness, patient engagement, and
appropriate human oversight. Patient-facing AI should help patients access
understandable information without encouraging inappropriate reliance on AI-generated
outputs when professional assessment is needed (AORN, 2026a, 2026b).

Recommendations for FDA Consideration
• Clear disclosure that AI is being used to generate or support responses.
Response: “Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper
and Request for Feedback,” Docket No. FDA-2026-N-7874
• Communication that is understandable and appropriate for the intended audience.
• Transparent communication of limitations, uncertainty, and intended use.
• Clearly defined pathways for escalation to licensed health care professionals.
• Privacy and security protections that safeguard patient information.
• Ongoing monitoring for unintended consequences and inappropriate reliance on
outputs.
• Evaluation of equity and potential disparities in access to care and patient outcomes.

Illustrative Example
For example, a patient-facing tool may provide preoperative instructions or answer
common perioperative questions, but organizations should define escalation pathways
when clinical judgment or individualized assessment is required (AORN, 2026a, 2026b).

Question 18: Premarket Evidence and Postmarket Monitoring
AORN Position
AORN supports robust postmarket monitoring as an essential component of lifecycle
management for AI-enabled technologies. Postmarket monitoring should complement,
rather than replace, appropriate premarket evidence demonstrating that an AI-enabled
technology is safe and effective for its intended use.

Rationale
AORN guidance emphasizes lifecycle governance, validation, monitoring, oversight, and
risk mitigation. As autonomy and potential patient harm increase, so should the strength of
premarket evidence and postmarket oversight. AORN identifies agentic AI as high risk
because of its potential for autonomous decision-making (AORN, 2026a, 2026b).

Recommendations for FDA Consideration
• Consider the severity of potential patient harm and the reversibility of actions
influenced by AI-generated outputs.
• Consider the level of autonomy exercised by the technology.
• Consider whether outputs are informational, action-directing, or integrated into patient
care processes.
• Consider the availability and effectiveness of human review mechanisms.
• Consider whether intended users can recognize and respond to erroneous, uncertain,
or out-of-scope outputs.

Response: “Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper
and Request for Feedback,” Docket No. FDA-2026-N-7874
• Consider the capacity of manufacturers and deploying organizations to detect,
investigate, communicate, and respond to performance concerns.
• Require evidence appropriate to the intended use, level of autonomy, and potential
consequence, with more intensive postmarket oversight as risk increases.

Illustrative Example
A tool that organizes information presents different risk than one that influences triage,
treatment decisions, or autonomous clinical actions. Technologies with greater autonomy
or potential consequences warrant stronger governance, validation, human oversight, and
monitoring (AORN, 2026a, 2026b).

Question 19: Postmarket Monitoring Approaches
AORN Position
AORN supports ongoing monitoring and reassessment of AI-enabled technologies
throughout their lifecycle. Monitoring activities should be incorporated into organizational
patient safety, quality, risk-management, and governance processes.

Rationale
AORN guidance supports ongoing evaluation because technology performance, data,
users, workflows, and clinical contexts may change. Monitoring should include patient
safety, workflow, personnel, privacy, security, equity, and trustworthiness (AORN, 2026a,
2026b).

Recommendations for FDA Consideration
• Evaluate clinical performance, reliability, validity, and fitness for the intended purpose.
• Monitor patient safety concerns, reported events, unintended consequences, patient
experience, and outcomes.
• Assess workflow effects, communication, usability, and end-user experience.
• Evaluate equity, fairness, bias, and potential disparities in access to care and patient
outcomes.
• Monitor privacy, security, data quality, and other data-governance concerns.
• Reassess performance after software updates, changes to underlying models or data,
or modifications to workflow integration.
• Provide formal, accessible, non-punitive processes for end users to report concerns.
• Define the scope and frequency of monitoring processes and findings and who has
authority to modify, restrict, or discontinue use.
Response: “Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper
and Request for Feedback,” Docket No. FDA-2026-N-7874
Illustrative Example
In perioperative settings, monitoring may identify workflow disruption, communication
problems, excessive alerting, inappropriate reliance, or performance changes. Findings
should be reviewed through governance processes and used to determine whether
education, workflow revision, additional validation, increased monitoring, restriction, or
discontinuation is warranted (AORN, 2026a).

Question 21: Stakeholder Roles in Postmarket Monitoring
AORN Position
AORN believes effective postmarket monitoring requires coordinated participation among
manufacturers, health care organizations, clinicians, professional societies, patients, and
other stakeholders. Shared participation should not diminish manufacturer accountability
for device safety and performance.

Rationale
AORN recommends interdisciplinary governance and engagement of health care
professionals throughout AI design, implementation, evaluation, and monitoring. Shared
participation should clarify, not dilute, accountability for safety and performance (AORN,
2026a, 2026b).

Recommendations for FDA Consideration
• Manufacturers should remain accountable for the safety, effectiveness, and
performance of their products and should communicate known limitations, safety
concerns, and changes that may affect use.
• Health care organizations should maintain interdisciplinary governance, evaluate local
implementation, monitor real-world use and workflow effects, and maintain reporting
and escalation pathways.
• Perioperative nurses and other clinicians should participate in evaluation and report
safety concerns, usability problems, workflow effects, and unintended consequences
identified in practice.
• Governance should include relevant clinical, informatics, information technology, datascience or statistical, legal or risk-management, ethics, quality, privacy, security, and
organizational leadership expertise, consistent with the needs of the technology and
setting.

Response: “Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper
and Request for Feedback,” Docket No. FDA-2026-N-7874
• Professional societies can develop evidence-based guidance, support education and
competency, articulate professional responsibilities, and communicate practice-based
concerns to regulators, manufacturers, and health care organizations.
• Roles, decision authority, and accountability should be defined so that shared
monitoring does not leave safety concerns without a responsible party or response
pathway.

Illustrative Example
Perioperative nurses are well positioned to identify workflow consequences,
communication problems, patient-safety concerns, and usability issues that may not be
captured by technical measures alone. Their role should extend across codesign,
implementation, evaluation, and governance, not only end use (AORN, 2026a, 2026b).

Question 26: Agentic AI Systems
AORN Position
AORN considers agentic AI a higher-risk category of AI-enabled technology because it can
autonomously plan, execute, and interact with external systems with reduced need for
continuous human intervention.

Rationale
AORN identifies agentic AI as high risk because of autonomous decision-making and
affirms that AI cannot replace the expertise, empathy, judgment, and accountability
required for clinical practice. Agentic systems therefore require clear boundaries, human
oversight, accountability, monitoring, and the ability to interrupt, override, or reverse
actions when feasible (AORN, 2026a, 2026b).

Recommendations for FDA Consideration
• Define intended-use boundaries and functions that the system is not permitted to
perform.
• Require human-oversight checkpoints before irreversible or high-consequence actions.
• Establish transparent accountability for actions proposed or initiated by the system.
• Provide mechanisms for qualified professionals to interrupt, override, or reverse
actions when feasible.
• Evaluate multi-step task performance, tool use, interaction with external systems, and
the possibility that errors may propagate across a sequence of actions.

Response: “Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper
and Request for Feedback,” Docket No. FDA-2026-N-7874
• Monitor for unintended autonomous actions, performance changes, and operation
outside the intended function.
• Apply governance, validation, education, monitoring, and risk-management
requirements proportional to the degree of autonomy and potential consequence.

Illustrative Example
A system that summarizes information requires different oversight than one that initiates
actions, executes multistep processes, or interacts with other systems. Greater autonomy
and more consequential actions warrant stronger oversight, validation, monitoring,
accountability, and intervention mechanisms (AORN, 2026a, 2026b).

Conclusion
AORN supports regulation that promotes innovation while maintaining patient safety,
accountability, and public trust. For GenAI-enabled medical devices, governance should
extend across the lifecycle, oversight should increase with autonomy and potential
consequences, and perioperative nurses and other frontline clinicians should remain
active participants in evaluation, implementation, monitoring, and decisions that affect
patient care.

References
Association of periOperative Registered Nurses. (2026a). Guideline for integration of
artificial intelligence. In Guidelines for perioperative practice. AORN, Inc.

Association of periOperative Registered Nurses. (2026b). AORN position statement on
artificial intelligence (AI) in perioperative nursing. AORN, Inc.

Response: “Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper
and Request for Feedback,” Docket No. FDA-2026-N-7874