AI can help people make life-and-death decisions, but it should not hold the final authority over them. In healthcare, public safety, or the use of lethal force, a human decision-maker needs the evidence, expertise, time, and power to choose differently—and must remain answerable for the outcome. A person who merely clicks “approve” on a decision they cannot meaningfully assess is not meaningful oversight.
What counts as AI making the decision?
The key question is not whether AI appears somewhere in the process. It is whether the system merely helps a person assess evidence, or effectively selects the outcome while the person has little practical ability to challenge it.
| AI’s role | What it may do | Who retains the decision |
|---|---|---|
| Information support | Organize records, flag patterns, or summarize evidence for review. | A human considers the output alongside other relevant information. |
| Recommendation | Suggest a diagnosis, course of care, or other action. | A human evaluates the recommendation, can reject it, and explains the decision. |
| Effective delegation | Determine an outcome that a person cannot realistically inspect, delay, or override. | Human authority is nominal; the system effectively decides. |
A human checkpoint is not enough by itself. Oversight is meaningful only if the person has relevant information and expertise, enough time to use them, and real authority to intervene. The organization deploying the system also needs to be identifiable and accountable.
What do the ethical guidelines say?
UNESCO’s 2021 Recommendation on the Ethics of Artificial Intelligence says that where decisions may have irreversible effects or involve life and death, “final human determination should apply.” It adds, “As a rule, life and death decisions should not be ceded to AI systems.” The Recommendation is a global ethical framework, not a universal statute that automatically creates enforceable legal duties in every country.
#1 Best Overall
In 2026, WHO’s policy discussion similarly described AI as something that should augment, not automate, human judgment. UN Secretary-General António Guterres also put the principle plainly in 2026 remarks: “in every high-stakes decision – in justice, in healthcare, in policing – machines can inform, but humans must decide – and answer.” These are important policy positions; they do not, on their own, establish the laws that apply in every jurisdiction or settle every legal question.
Where can AI help—and where can it go wrong?
Healthcare
WHO identifies potential uses across diagnosis and screening support, clinical care, research and drug development, disease surveillance, outbreak response, and health-system management. AI may also help extend services to rural or underserved communities where access to health professionals is limited. These are reasons to assess carefully governed assistance, not to assume that a system is safe or effective for every patient or setting.
WHO also warns that health-related large language model answers can sound authoritative while being seriously wrong. A fluent explanation is not proof that a recommendation is accurate. Another concern is whether a system’s data and evaluation reflect the people who will be affected: WHO cautions that tools developed mainly with data from high-income countries may not perform well in low- and middle-income settings. Errors that are uneven across populations can worsen existing inequities.
WHO Director-General Dr Tedros Adhanom Ghebreyesus said in 2021: “Like all new technology, artificial intelligence holds enormous potential for improving the health of millions of people around the world, but like all technology it can also be misused and cause harm.” WHO advises careful risk assessment, expert supervision, and evidence of benefit before widespread routine use. Adoption should not be treated as a substitute for investments needed to achieve universal health coverage.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchPC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Rank #3
Lethal force
AI used in decisions about lethal force raises both ethical and legal stakes. In a statement cited here, the European Union argues that people should retain control over lethal force and remain accountable, linking that position to principles of international humanitarian law including distinction, proportionality, and precautions. That is the EU’s position; one statement is not a complete account of international law, state practice, or the rules applicable to a particular operation.
Why should a person retain final authority?
- Responsibility and redress: When a decision causes harm, an affected person needs to know which organization or person is answerable and how to challenge the decision or seek a remedy. UNESCO emphasizes attributable responsibility and redress; WHO calls for accountability and mechanisms for people to question algorithm-based decisions.
- Context and values: High-stakes choices may involve uncertainty, competing values, and circumstances that available data do not capture. This is a reasoned ethical argument for human determination, not a claim that people always make better decisions or that every human judgment is fair.
- Unequal errors: Data and evaluations that do not represent the affected population can produce misleading outputs or unequal performance. A system that works in one population or setting cannot simply be assumed to work in another.
- Meaningful control: A person cannot take responsibility for a choice if the process hides the relevant evidence, gives no time to examine it, or makes an override impossible in practice.
How to assess safeguards in a high-stakes AI system
Before accepting claims that an AI tool is safe for a consequential use, ask how the system fits into the full decision process—not just how it performs in a demonstration.
- Define the decision. What exact action is the system meant to support, and what evidence shows safety, accuracy, and benefit for that use?
- Check whose data count. Were the data and evaluations representative of the people and settings affected? Are limitations made clear?
- Test the human role. Can the decision-maker understand the system’s limits, see relevant evidence, and intervene in time? Is an override genuinely available?
- Protect information and explain responsibility. Are privacy and security addressed, is the process transparent enough for appropriate scrutiny, and is an accountable organization responsible for deployment?
- Provide a route to challenge and remedy. Can someone affected by an adverse decision question it and obtain meaningful redress?
- Review deployment over time. Are readiness reviews and impact assessments completed before use, with ongoing monitoring and multidisciplinary oversight afterward?
These safeguards can reduce avoidable risk, but they do not prove that transferring final life-and-death authority to AI is safe in every context. WHO’s 2026 policy discussion describes practical measures such as readiness reviews, impact assessment, human verification, decision gateways, and multidisciplinary oversight; the underlying principle remains that AI should strengthen human deliberation rather than replace it.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
