AMA: Keep Physicians at the Center of AI - With Real Oversight, Secure Data, and Ongoing Training
The American Medical Association is urging a careful, physician-led approach to AI in healthcare. The group explicitly frames AI as "augmented intelligence," meant to enhance clinical judgment - not replace it.
The message to policymakers and health leaders is clear: prioritize education, physician oversight, and data security. Patient safety and clinical accuracy come first, and that means physicians stay in the loop at every step.
What the AMA Recommends
- Physician-led decisions: Clinicians should guide AI selection, validation, and use, and review outputs to protect accuracy and safety. Expertise cannot be substituted.
- End-to-end clinical involvement: Physicians should be full partners across the AI lifecycle - from design and testing to integration into workflows.
- Coordinated oversight: Federal agencies should align to create a coherent oversight system. Fragmented or duplicative rules slow progress and confuse teams.
- Secure, unbiased data: Strong deidentification, consent safeguards, and transparency to patients and physicians about how data is used and protected.
- Upskilling the workforce: Education in medical school and CME for practicing clinicians so teams can properly assess and apply AI tools.
Why This Matters for Health Systems
Done right, AI can support patient-centered care, improve outcomes, and lower costs. Done poorly, it introduces clinical and operational risk that compounds over time.
Most organizations are already experimenting. Eighty-eight percent of health systems report internal AI use, yet only 18% have mature governance and a full strategy. Meanwhile, 71% have pilots or live tools in finance, revenue cycle, or clinical areas - often without the guardrails to match.
Data Privacy, Bias, and Trust
Trust hinges on the data. The AMA calls for strong deidentification, consent, and clear disclosures to patients and clinicians. Systems should continuously monitor for bias and document data provenance.
If your AI depends on data you can't explain or defend, it will fail audits, lose clinician confidence, and potentially harm patients. Treat data governance as clinical quality work, not an IT task alone.
Upskilling Physicians (and Teams)
The AMA emphasizes ongoing training so physicians can evaluate AI claims, interpret outputs, and flag risks. That includes medical education and CME, plus practical, workflow-level training for care teams.
The AMA's new Center for Digital Health and AI will support education, training, and policy collaboration. For context on AMA positions around "augmented intelligence," see the AMA's AI resources here. The association also submitted its views to the Senate HELP Committee, which you can explore here.
If you're building skills for clinical and operational teams, you can also explore curated AI courses by role at Complete AI Training.
Practical Moves You Can Make Now
- Stand up real governance: Establish a cross-functional AI council (clinical leaders, quality, compliance, security, informatics, and frontline clinicians). Give it authority over intake, validation, deployment, and monitoring.
- Define clinical validation standards: Require indication-specific evidence, bias and safety checks, and clear performance thresholds before go-live. Document who reviews and how often.
- Lock in data safeguards: Enforce deidentification, consent management, data minimization, and role-based access. Track lineage and audit usage.
- Train for real workflows: Build short, role-based training for physicians, nurses, and support staff. Teach failure modes, proper prompts (if applicable), and when to override the tool.
- Monitor post-deployment: Set up continuous performance, drift, and bias monitoring. Create a rapid rollback path when metrics slip or safety signals appear.
- Invest in infrastructure: Strengthen telehealth, privacy/security controls, and interoperability so AI tools integrate cleanly without adding friction.
The Bottom Line
AI can help clinicians deliver better, more efficient care - but only under physician oversight, with secure data, clear governance, and ongoing training. That's the AMA's stance, and it's a practical playbook for health systems already experimenting with AI.
Put physicians at the helm. Build guardrails before scale. Educate your teams. That combination earns trust and keeps patients safe.
Your membership also unlocks:
