On October 30, President Joe Biden issued an Executive Order on the “Safe Secure and Trustworthy Development and Use of Artificial Intelligence,” which includes provisions that will potentially have a broad impact on healthcare as it seeks to ensure the “responsible deployment and use of AI and AI-enabled technologies” in healthcare as well as other areas.  

This includes overseeing the following areas, which will fall under the responsibility of an AI Task Force to be formed within the Department of Health and Human Services (HHS): 

• The development, maintenance, and use of AI-enabled technologies 
• The monitoring of long-term safety and the real-world performance of AI-enabled technologies 
• The incorporation of standards in AI-related software development cycles; and 
• Comprehensive documentation efforts that can assist with the use of AI in local settings  

Such oversight by the AI Task Force would include developing a strategic plan that includes policies and frameworks and as will be seen in more detail below, the Executive Order is also focused on quality maintenance when it comes to AI-enabled technologies and seeks to develop an “AI assurance policy” as to such efforts.  

 SAFETY 

The Executive Order places a large focus on AI-related safety generally, and in the healthcare setting, the focus includes a special emphasis on the protection of patients. To this end the Order creates an AI Safety Program. 

AI Safety Program 

Found in paragraph 8(b)(iv), the AI Safety Program is to address harms related to unsafe healthcare practices involving AI-enabled technologies as follows: 

• It is to increase the ability to identify and capture clinical errors by establishing a common framework for the deployment of AI in healthcare settings 
• It is to create a central tracking repository for incidents that cause harm 
• It is to develop guidelines or best practices to avoid harm based on captured data and generated evidence 
• It is to make guidelines widely available to stakeholders such as healthcare providers  

Other Provisions 

A provision not part of the AI Safety Program that is found in paragraph 8(b)(i)(B) of the Executive Order provides for the monitoring of AI-enabled technologies in terms of both long-term safety and real-world performance. 

Not specific to healthcare is the requirement to develop guidance on AI red-team testing, or ‘red-teaming,’ defined in paragraph 3(d) as structured testing efforts to find flaws and vulnerabilities such as harmful or discriminatory outputs in AI systems, often in collaboration with developers of AI.  

Additionally, companies developing or showing an intent to develop potential dual-use foundation models of AI must share the items specified in paragraph 4.2(a)(i) of the Executive Order with the Federal Government, which includes information, reports, and records concerning such things as the training and developing of such models and the results of any developed model’s performance in relevant AI red-team testing.  

A dual-use foundation model is defined in paragraph 3(k) as an AI model that is trained on broad data; generally uses self-supervision; contains at least tens of billions of parameters; is applicable across a wide range of contexts; and that exhibits, or could be easily modified to exhibit, high levels of performance at tasks that pose a serious risk to security, national economic security, national public health or safety, or any combination of those matters, with more specifics provided therein. 

PRIVACY 

A large focus is also placed on AI-related privacy issues and although the protection of personally identifiable information is currently covered by HIPAA, paragraph 8(b)(i)(D) of the Executive Order seeks to go further and includes the protection of such information during the software-development lifecycle by incorporating privacy and security standards at this stage. Such measures are also to include a focus on AI-enhanced cybersecurity threats.  

Privacy and nondiscrimination are further covered in paragraph 8(b)(iii) which adds an AI-specific focus to existing laws.  

Additionally, paragraph 8(a) provides ‘encouragement’ to independent regulatory agencies to “consider rulemaking, as well as emphasizing or clarifying where existing regulations apply to AI,” and to consider “clarifying the responsibility of regulated entities to conduct due diligence on and monitor any third-party AI services they use.”  

However, it can be noted that this approach has recently brought about legal action in opposition to it, specifically in the form of the American Hospital Association (AHA)’s lawsuit in the U.S. District Court for the Northern District of Texas with regards to HHS’s Bulletin from December of 2022 concerning the “Use of Online Tracking Technologies by HIPAA Covered Entities and Business Associates.” In its November 2 Complaint, the AHA alleges that prohibitions on such use in the Bulletin extend too far and constitute “overreach by the federal bureaucracy,” highlighting the complex nature of issues surrounding privacy.   

QUALITY 

Paragraph 8(b)(ii) calls for the development of an AI assurance policy to make sure that proper levels of quality are maintained by AI-enabled technologies in healthcare.  

Specific objectives include evaluation of the “important aspects of the performance of AI-enabled healthcare tools,” as well as the infrastructure needs for enabling pre-market assessment and post-market oversight of AI-enabled technology algorithmic system performance against real-world data.  

ADDITIONAL AREAS OF FOCUS 

Drug Development  

Paragraph 8(b)(v) focuses on regulating the use of AI-enabled tools in drug development. As elsewhere the Executive Order seeks a comprehensive approach, ordering HHS to focus on each phase of drug development and define “objectives, goals, and high-level principles,” needed for effective oversight. 

Equity Principles 

The Biden Executive Order does not neglect the incorporation of equity principles in AI-enabled technologies.  

Paragraph 8(b)(i)(C) requires the use of disaggregated data on affected populations and representative population data sets when developing new models. It also calls for the monitoring of performance against discrimination and bias in currently existing models. 

Additionally, although not specific to the health and human services sector, the entirety of Section 7 of the Executive Order is devoted to equity and civil rights concerns. 

Innovation 

In addition to regulating AI-enabled technologies, the Biden Executive Order also seeks to promote innovation, stating it is not only seeking to ‘manage the risk’ but also to ‘seize the promise of AI.’ 

Section 5.2(b) calls for supporting training for AI scientists and researchers to increase AI talent generally, and when it comes to healthcare, Section 5.2(e) calls for additional focus when it comes to grantmaking and other awards for the development of such things as AI tools for clinical care.

Categories: AI, Cybersecurity, DATA PROTECTION & PRIVACY