AAIH Response to White House Executive Order on AI
Rafael Rosengarten, PhD, Co-founder and CEO, Genialis & Co-founder and Board Director, The Alliance for AI in Healthcare
Elaine Hamm, PhD, Executive Director, The Alliance for AI in Healthcare
& the Board of Directors and Members of AAIH
On October 30, the Biden Administration released a sweeping Executive Order (EO) aimed at framing its approach to regulating artificial intelligence (AI). While healthcare was expressly mentioned as one of the sectors covered by the EO, the White House’s declaration is aimed at a “society-wide effort that includes government, the private sector, academia, and civil society.” The Alliance for AI in Healthcare (AAIH) welcomes this decisive policy direction. Further, considering AI’s transformative capacities, we stress that the healthcare industry, with its well-established frameworks for patient safety, privacy, and other safeguards, is uniquely poised to integrate and expand these principles in the realm of AI in Healthcare and beyond.
The AAIH was inaugurated in January 2019 as a not-for-profit global advocacy organization for the responsible adoption of AI to improve healthcare equity and outcomes. In its EO, the White House stated, “AI reflects the principles of the people who build it, the people who use it, and the data upon which it is built.” This sentiment resonates deeply with our organization. For four years we have assembled leaders from all areas of healthcare to contribute their insights, voices, and efforts to this vital endeavor. Our significant efforts range from educational white papers and webinars, gatherings for thought-leaders, recommendations on technology and data standards, as well as regulatory guidance to both Congress and the FDA.
We would like to take this opportunity to address the eight principles that the White House has put forward. Importantly, the healthcare sector has done a commendable job already in setting an example for how these principles can be implemented and upheld.
Artificial Intelligence must be safe and secure.
Any AI-based device or invention that requires FDA security already must meet the standard set by the EO, namely: “robust, reliable, repeatable, and standardized evaluations of AI systems, as well as policies, institutions, and, as appropriate, other mechanisms to test, understand, and mitigate risks from these systems before they are put to use.” To this end, the FDA has released guidelines on both Good Machine Learning Practice (GMLP) and AI within Software as a Medical Device (SaMD).
In terms of security concerns, even though there are often calls to revamp HIPAA, healthcare data and IT systems rank among the most safeguarded digital categories. Compliance with standards such as HIPAA (and GDPR in Europe), ISO 9001, ISO 27001, and ISO 13485 are foundational requirements for operating in this domain.
Responsible innovation, competition, and collaboration “requires stopping unlawful collusion and addressing risks from dominant firms’ use of key assets such as semiconductors, computing power, cloud storage, and data to disadvantage competitors, and it requires supporting a marketplace that harnesses the benefits of AI to provide new opportunities for small businesses, workers, and entrepreneurs”
While it may seem idealistic, history has demonstrated that innovation is often spearheaded by entrepreneurs and emerging startups in the realm of technological disruption. Indeed, our members aim to ensure vigorous competition and equitable practices. Large companies have an important role to play in commoditizing the infrastructure on which we can most equitably deliver and distribute healthcare innovations. We have observed that the complexities and uniqueness of healthcare challenges are such that innovative solutions almost invariably come from agile inventors rather than megalithic incumbents.
Commitment to supporting American workers
Healthcare practitioners, by and large, are highly trained and mission-driven. The goal of AI systems in this space typically is to liberate the healthcare workforce from tedious and error-prone activities and give back more time to skilled work and human interactions. Whether it's assisting doctors in spending more meaningful time with their patients, aiding researchers in discovering breakthrough treatments faster, or facilitating quicker and safer drug market entries, AI in healthcare presents an invaluable opportunity to enhance the impact of human expertise. We believe that AI will make healthcare more human, not less.
Consistent with advancing equity and civil rights
AI can be only as good as the data that feeds it. Thus data equity and healthcare access remain important concerns. But these concerns are also at the heart of healthcare and medicine as a whole. Systemic change that addresses inequality based on gender, wealth, geography, ethnic background, etc is required. Addressing these issues will go a long way to mitigating risks that AI will propagate bias and enshrine inequality. In the meantime, the AAIH has, from its inception, called for practitioners to acknowledge and address bias in all its forms, from diversity in datasets to pervasive human and social bias, and ensuring that access to such data is as open and available as possible.
Consumer protection
Key to consumer protection is consumer and provider education. The AAIH maintains market education as a priority. We also note a difference between healthcare AI algorithms, devices, and products that fall under the watch of the FDA (or EMA in Europe) versus other products that enter the public domain or consumer markets without such scrutiny. For example, there is a significant difference between an AI algorithm embedded in a certified diagnostic device for directing cancer treatment and an online chatbot that merely allows individuals to speculate about their health issue. We argue that AI in the former case is sufficiently regulated, while the latter arena is ripe for guardrails. We believe that by understanding and internalizing this critical distinction, the White House and other policymakers can pave the way for more effective and responsive AI regulations in healthcare and beyond.
Privacy and civil liberties must be protected
We concur, highlighting that the healthcare sector already has a strong foundation in place with existing data protections and regulations against discrimination. The path to tackling bias in AI begins with equitable healthcare access. As our nation works to make healthcare more accessible, the risks associated with data bias will decrease. As more derived-data products are generated from AI that is trained on healthcare data, more efforts to safeguard underlying patient data should continue to be considered. When patients opt in to sharing their data for research purposes, for instance, the benefit could be new AI-driven diagnostics or treatments that can benefit all patients. Patients that share data should have absolute confidence that their privacy is not at risk, that their identity or data will not be revealed or reverse-engineered from the AI models trained on it.
Manage the risks from the Federal Government’s own use of AI
From its inception, the AAIH has actively sought to engage with government agencies. We are enthusiastic and ready to serve as a resource of expert advisors as government entities define their own operational guidelines.
Global leadership, international allies
We agree that the US leadership role is stronger when complemented by international allies. Powerful nations and large economies have a duty to the global community to consider the needs of the less powerful and well-represented. We must work together to enact the broadest definitions of equality and inclusion while preventing exploitation and exclusion.
In sum, the healthcare sector has already taken up the challenge of responsible adoption of AI. The regulatory agencies and policymakers have safeguards in place that protect against many of the concerns enumerated in the current EO. Moreover, these same entities consistently have shown themselves to be thoughtful, curious and engaged when it comes to learning about and adapting to this fast-moving area of technology. We encourage the federal government to continue to include stakeholders from across healthcare—from patient advocacy groups to practicing physicians, drug developers to provider systems—in its quest to usher in the benefits while mitigating the risks posed by AI. While there is still much for us to learn, we also possess extensive knowledge that can help others navigate the responsible adoption of AI.
About AAIH
The Alliance for Artificial Intelligence in Healthcare (AAIH) is an international advocacy organization dedicated to the responsible adoption and application of AI/ML in healthcare. Our members represent all healthcare sectors, including drug R&D, digital health, medical devices, healthcare delivery, data science & technology providers with a shared goal of improving the state of healthcare globally through AI-enabled solutions. For more information, please visit www.theaaih.org
