Healthcare adoption still trails other industries in AI use, and public trust remains limited, which reinforces the need for strategic leadership, clear guardrails, and responsible implementation.
There’s more attention, more skepticism and more pressure to get AI right, especially knowing these applications can impact patient lives. As AI becomes more integrated into drug development and clinical workflows, leaders across healthcare and life sciences are therefore expected to demonstrate ethical stewardship and technical competence to instill trust and protect patients.
That’s why ethical AI and trustworthy AI frameworks should be table stakes for healthcare and life sciences organizations seeking to innovate with AI. These distinct, but interconnected, frameworks contribute to a defensible AI strategy — one that ensures all AI-related actions and decisions can be explained and justified to key stakeholders. Together, they allow teams to answer two critical, pressure-tested questions: what risks are associated with this AI application, and are we capable of mitigating them?
The differences between ethical AI and trustworthy AI
The distinction between “ethical AI” and “trustworthy AI” is meaningful for pharmaceutical executives, healthcare providers and regulators who influence how AI is developed and deployed. While the concepts can overlap, it’s useful to consider them separately when setting strategy, guardrails and evidence expectations.
Ethical AI and trustworthy AI are operational frameworks that shape how organizations govern, deploy and earn trust for AI solutions. Their interplay and sequence vary depending on context and organizational priorities. In practice, organizations may address ethical and trustworthy AI concurrently, using trustworthy AI to define what can be engineered and evaluated while ethical considerations guide oversight across the lifecycle.
Here’s the main difference: ethical AI examines what should be done by reflecting an organization’s leadership posture, through values, norms and fairness. Ethical frameworks are specialized and balance policy, technology, and business needs. Because many ethical considerations involve nuance and tradeoffs, frameworks should be adaptable as new information or context emerges.
For example, when using social determinants of health (SDoH) to guide resource allocation using AI for public health, are we relying on community‑level indicators that respect context, or individual‑level signals that could feel intrusive or unfair? In clinical trial recruitment, is it acceptable to optimize AI for predicted completion if that reduces population representativeness, and where is the ethically defensible balance?
Trustworthy AI examines what can be done. These frameworks focus on technical requirements (what needs to be built and engineered) and emphasize demonstrable compliance, reliability, safety, security, transparency and explainability. Trustworthy AI also accounts for binding legislation and recognized frameworks, such as the EU AI Act (legislation) and the NIST AI Risk Management Framework (guidance), which advise organizations on responsible AI practices shaping how the principles of trustworthy AI are evidenced.
Following the previous examples, if a public health department does use SDoH (perhaps at a community level), trustworthy practice means engineering for patient safety and data security, including technical controls and access patterns needed to satisfy AI and data protection laws, meeting relevant standards, and producing verifiable evidence of those controls. For trial recruitment, balancing population diversity with trial completion likelihood in recruitment efforts, the focus will include accurate completion predictions across all subpopulations. The recruitment strategy also needs to remain clinically sound and defensible to regulators like the FDA or EMA.
A cohesive model
Understanding the differences between ethical and trustworthy AI is one thing. Putting them into practice is another. The question becomes: how do we operationalize both frameworks simultaneously?
Ethical AI and trustworthy AI should weave together under cohesive strategic oversight, providing clear guardrails, roles and accountabilities that adapt by use case. Drug developers exploring AI use for adverse event detection will have different risks and different workflows than a hospital using AI for clinical documentation. But both need accountability structures in place that prove AI use is ethical, trustworthy, responsible and defensible.
Put ethics and trust to work
Building cohesive oversight and clear direction requires deliberate collaboration. Regulations and societal expectations are evolving across industries and regions. Leaders can de-risk progress by setting strategic guardrails early, measuring safety and security outcomes, and updating evidence as standards mature.
To meet heightened scrutiny and demonstrate strategic oversight, transparency and accountability, organizations can develop their own frameworks that balance operational efficiency with patient safety, security, and ethical guardrails. Here are some considerations for getting started:
- Establish AI as a strategic priority: Intent, impact and technical design principles need to be defined at the outset so that purpose, patient safety, and security are built in rather than bolted on. Setting an ethics‑led strategic posture (the why) alongside trustworthy engineering standards (the what) gives teams clear guardrails and direction from day one. Cross‑disciplinary leadership (including clinical, product, AI science, security, privacy, legal) co‑owns this posture and evolves it as evidence and regulations mature. Working this way reduces late‑stage surprises and aligns pilots and development with the culture and expectations of healthcare and life sciences. This helps reduce later‑stage challenges, supports implementation, and lowers operational risk.
- Formalize guardrails and direction: Strategic oversight supports innovation by creating clarity about how AI is reviewed and applied. In healthcare and life sciences, a formal ethics and trust oversight board that brings together senior leaders and technical experts can provide a consistent way to assess activities, surface evidence gaps, and resolve questions about accountability. This structure helps teams identify high‑risk or ambiguous use cases early and apply the right level of scrutiny with patient safety and security as first principles. When oversight runs in parallel with implementation workstreams, organizations address regulatory duties, risk, and ethical questions as part of the build, so responsible innovation moves faster and with confidence.
- Leverage a tiered risk approach: A healthcare or life sciences organization, supported by privacy and compliance experts in a review board with executive oversight, should assess AI use cases through a tiered risk approach — one that considers the type of AI as well as its position in the product lifecycle and proximity to patient lives. This adaptive approach scales the evaluation of AI risk and credibility based on need (where guardrails have already been implemented for many applications). For instance, literature mining and scenario modeling are lower-risk AI applications used in early-stage research, which is also considered lower risk. Whereas higher‑risk applications, such as large language models used in clinical decision support, require more rigorous evaluation because of their complexity and proximity to patient care. Patient safety is the imperative and the lens through which technical and business trade-offs are made.
- Build cultural readiness: This is a core part of deploying AI effectively because teams need to understand how the technology fits into their work and how decisions are being made. AI programs gain traction when leaders explain how applications support ethical commitments, patient-safety outcomes, security standards, and business objectives. Leaders will also need to address questions about workflow changes, regulatory expectations or perceived job impacts with clarity. Providing structured change management support helps people see how AI can contribute positively to operational goals and patient outcomes, which builds confidence and enables adoption across the organization.
The need for careful implementation is constant because AI in healthcare and life sciences inevitably influences patients and their care. With strategic oversight, clear guardrails, patient‑safety measures, and secure engineering, organizations can apply AI responsibly, earning the confidence of clinicians, regulators, and, most importantly, patients.
Photo: Pakorn Supajitsoontorn, Getty Images
Luk Arbuckle is Global AI Practice Leader and Chief Methodologist at IQVIA Applied AI Science, driving innovation in data and AI with a focus on ethics, privacy, and regulatory compliance. A recognized thought leader, Luk advises global authorities and publishes widely on AI governance and data protection.
This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.
