AMA issues principles for guiding AI development and use

The American Medical Association (AMA) has issued a set of principles to guide the development and use of "augmented intelligence" (AI) in healthcare. Recognizing the potential benefits and ethical considerations of AI, the AMA's principles cover various aspects:

Oversight: The association encourages a government-wide approach for policy implementation to mitigate risks associated with healthcare AI, while also acknowledging the role of non-government entities in appropriate oversight.

Transparency: The design, development, and deployment of AI processes should be mandated by law to include key characteristics and information. Transparency is deemed essential to establish trust among patients and physicians.

Disclosure and Documentation: The statement calls for appropriate disclosure and documentation when AI directly impacts patient care, medical decision-making, access to care, communication, or the medical record.

Generative AI: Health care organizations are encouraged to develop and adopt policies that anticipate and minimize potential negative effects of generative AI, with these policies in place prior to adoption.

Privacy and Security: AI developers should design systems with privacy in mind, implementing safeguards to assure responsible handling of patients' personal information. Strengthening AI systems against cybersecurity threats is crucial.

Bias Mitigation: Proactive identification and mitigation of bias in AI algorithms are emphasized to promote equitable health care outcomes and a fair, inclusive healthcare system.

Liability: The AMA advocates for limited physician liability for the use of AI-enabled technologies, aligning with current legal approaches to medical liability.

Additionally, the statement urges payors not to use automated decision-making systems in ways that reduce access to needed care or withhold care from specific groups, emphasizing the importance of not overriding clinical judgment and maintaining human review of individual circumstances. The principles aim to maximize AI's benefits while minimizing potential harm to patients and clinicians.