Ethical AI in Healthcare and Senior Care: A Human-Centered Guide for Leaders

Ethical AI in senior care means using artificial intelligence in ways that protect older adults' dignity, autonomy and safety, give caregivers more time for people rather than less, and keep humans accountable for every decision that matters. In practice, it rests on four habits: consent, inclusion, transparency and governance that leaders own.

What does "ethical AI" actually mean in healthcare and senior care?

Ethical AI in care means the technology serves the person receiving care and the person giving it, and never quietly replaces judgment, consent or relationship. It is a standard for decisions, not a feature a vendor can switch on.

I came to senior housing from a career in technology, so I understand the pull of a well-designed tool. After about fifteen years in the sector, I also understand what can't be automated. Trust between a resident and a caregiver. A family's sense that someone is paying attention. The judgment of a nurse who notices that something is "off" before any chart shows it.

That is why I find one test more useful than any glossary. When a new AI tool arrives, ask: whose time does it give back, and whose voice does it take away? A tool that cuts documentation so a caregiver can sit with a resident for ten extra minutes is moving in the right direction. A tool that makes decisions about a resident's day without that resident, or their family, understanding why is not, however accurate it claims to be.

Older adults and people living with dementia are among the people least likely to be asked what they think of the systems now being built around them. Ethical AI starts by changing that.

What are the WHO's six principles for AI in health?

In 2021, the World Health Organization published six guiding principles for AI in health. They are the clearest starting point I know for any leadership team.

Protecting human autonomy — Do residents and families stay in control of decisions about their care?

Promoting human well-being and safety and the public interest — Have we tested for harm, not only for benefit?

Ensuring transparency, explainability and intelligibility — Could a caregiver explain what this tool does to a resident's daughter?

Fostering responsibility and accountability — Is a named person accountable when the tool gets it wrong?

Ensuring inclusiveness and equity — Does it work as well for an 88-year-old as for a 58-year-old?

Promoting AI that is responsive and sustainable — Will we keep reviewing it, and retire it if it stops serving people?

I like that autonomy comes first. In senior care, autonomy is fragile. It erodes slowly, often through well-meant decisions made for someone rather than with them. AI can speed that erosion or help reverse it, depending on how we deploy it.

Which of these six could your organization not answer confidently today?

How can AI be ageist, and how do we prevent it?

AI becomes ageist when older adults are underrepresented in its data, its design or its testing, so the system works less well for them or repeats old stereotypes. Prevention means designing with older people, not just for them.

The WHO's 2022 policy brief, Ageism in artificial intelligence for health, warns that "the data used can be unrepresentative of older people or skewed by past ageist stereotypes, prejudice or discrimination." In other words, a model trained on yesterday's assumptions can automate them at scale.

The brief sets out eight considerations. Four of them translate directly into leadership choices:

• Participatory design "by and with older people," not focus groups held after the product ships.
• Age-diverse data science teams, so someone in the room notices what is missing.
• Age-inclusive data collection, so the oldest residents are represented in the data rather than excluded as "outliers."
• Rights to consent and contest, so older people can say no and challenge decisions made about them.

Ageism in AI is rarely deliberate. It usually comes from absence: nobody over seventy tested the voice interface, nobody asked how a person living with dementia experiences a camera in their room. I've seen in senior living how much a resident's sense of home depends on small dignities. Technology should add to them.

What do the EU AI Act and the NIST AI Risk Management Framework ask of leaders?

Both ask leaders to know where AI is used, judge its risk and keep humans accountable. The EU AI Act is binding law with staged deadlines. The NIST framework is voluntary guidance that many US organizations use to structure their governance.

The EU AI Act. According to the European Commission, the Act entered into force on 1 August 2024. Prohibitions and AI literacy obligations applied from 2 February 2025, and the Act became applicable on 2 August 2026, with some exceptions. Under the AI Omnibus changes, the Commission states that deadlines for many high-risk systems have been extended to 2 December 2027, and to 2 August 2028 for AI embedded in regulated products such as medical devices. The Act sorts AI into risk tiers: unacceptable, high, transparency (limited) and minimal. Timelines have already moved once, so I'd treat these dates as a current snapshot and check the Commission's page before relying on them.

The NIST AI RMF. NIST released its AI Risk Management Framework 1.0 on January 26, 2023, for voluntary use. It is built around four functions:

1. Govern: set roles, policies and a culture of accountability.
2. Map: understand where and how AI is used, and who it affects.
3. Measure: assess performance, bias and risk over time.
4. Manage: act on what you find, including stopping use.

For leaders outside Europe, the lesson is not to wait. Even where the law doesn't yet reach you, families, regulators and your own staff will expect you to show your thinking.

How should a healthcare or senior-living organization govern AI?

Govern AI the way you govern clinical quality: with a named owner, a simple inventory, clear risk tiers and regular review that includes the people affected. Good governance is less about paperwork and more about who is invited to the table.

A starting structure I'd recommend:

• One accountable executive, with authority to pause a tool.
• An AI inventory. Many organizations are surprised by how much AI is already embedded in software they bought for other reasons.
• A small review group that includes a frontline caregiver, a clinician, someone from compliance and, wherever possible, a resident or family voice.
• Risk tiers. A scheduling assistant and a fall-prediction model deserve very different scrutiny.
• A consent practice written in plain language, revisited when a resident's capacity or wishes change.
• AI literacy for staff, so caregivers can explain tools honestly and raise concerns without fear.

Frontline staff deserve particular attention. They will be the ones explaining a tool to a worried family at the end of a long shift. If they don't trust it, residents won't either.

Who in your organization has the authority to switch off an AI tool, and do they know it?

What questions should leaders ask an AI vendor before signing?

Ask questions that reveal how the tool was built, who it was tested on and who carries responsibility when it fails. A vendor's answers to these will tell you more than any demo.

1. How were older adults, including people living with dementia, represented in your training data and testing?
2. Were older people or care partners involved in designing this product?
3. What does the tool do with resident data, where is it stored and who can access it?
4. Can a caregiver understand and explain why the system made a given recommendation?
5. What happens when the tool is wrong, and how will we find out?
6. How does this give time back to caregivers, and how will we measure that?
7. How will you support us in meeting the EU AI Act or NIST-aligned governance, where relevant?
8. Can a resident or family opt out without losing quality of care?

If a vendor can't answer the first two, I'd slow down. Not because the tool is necessarily bad, but because nobody has yet checked whether it works for the people you serve.

If your board or leadership team is weighing how to bring AI into care responsibly, I'd be glad to help through my keynote, Ethical Artificial Intelligence With a Human Heart, or the Navigating Change workshop. You can reach me through the Book Ki page on siadatan.com.