As a part of the American AI Initiative, a new draft memorandum proposed by the White House yesterday identifies policy that will guide regulatory and non-regulatory oversight of AI applications, including medical device AI.
The American AI Initiative seeks to:
- Ensure that ingenuity is not restricted by over-regulation
- Certify that AI is used for public benefit
- Prevent AI from infringing on civil liberties
Federal agencies will be required to adhere to the principles laid out in the memorandum when proposing regulations for AI technologies.
The document’s scope only applies to private companies – government use of AI remains outside the scope of the recent memorandum.
It’s no secret that the application of AI raises unique ethical questions. We’ve developed entire genres of cinema, literature, and art devoted to doomsday scenarios where a sinister AI subjugates humanity to a brutal mechanical dictatorship.
However, this memorandum only applies to weak AI. Weak AI, while exceeding conventional computing practices to utilize data sets to learn and perform tasks, is not strong AI. The memorandum defines strong AI as computers that exhibit consciousness and may perform tasks better than a human agent.
The White House outlined key principles agencies need to adhere to when proposing medical device AI regulations.
These principles are:
- Public Trust in AI
The White House predicts that AI will have a net positive economic and social impact in healthcare. However, it recognizes that it could present privacy risks if used inappropriately, so it understands the importance of enforcing ethical uses to obtain public trust.
- Public Participation
The White House believes that public participation will improve the accountability of the government to ensure that AI does not infringe upon civil rights. It encourages regulatory agencies to inform the public and promote awareness of AI standards.
- Scientific Integrity and Information Quality
The White House believes that regulation should be based on scientific and technical information, and that information pertaining to the strengths and weaknesses of AI should be transparent.
- Risk Assessment and Management
Regulatory approaches to AI needs to be based on a spectrum of risk and reward, and it is not necessary to avoid every foreseeable risk if there is substantial reward.
- Benefits and Costs
The benefits and costs of a specific regulation will be considered when authoring. In other words, regulators should consider the full economic cost before creating a regulation.
Agencies should adopt flexible approaches that can adapt to technological changes in order to avoid stifling innovation.
- Fairness and Non-Discrimination
While AI has the potential to reduce discrimination caused by subjectivity, it does have the possibility of producing discriminatory outcomes from real-world bias. Agencies should ensure that AI applications reduce unfair, unlawful, or unintended discrimination.
- Disclosure and Transparency
The White House believes that transparency in rule-making will help AI gain public trust and adoption.
- Safety and Security
Agencies should create rules that ensure AI systems are secure, safe, and function as intended.
- Interagency coordination
The White House’s approach to AI involves the entire government, so agencies are encouraged to share experiences to ensure consistency.
We’ll have more on this breaking story as it pertains to medical device AI. You can read the entire memorandum here.