—
Like me, I’m sure you’ve seen several headlines claiming that “Artificial intelligence is now at PhD level,” or that it has a “higher IQ than any human alive.” Putting aside the vacuous and hollow nature of such statements, AI’s capabilities remain awe inspiring. It certainly can analyze, spot patterns, recommend, and persuade with unnerving confidence and at speeds no human could ever replicate. But its most humanlike attribute might indeed be that it can also kill.
On the evening of March 18, 2018, a self-driving vehicle in Tempe, Arizona struck and killed a pedestrian, Elaine Herzberg. The system detected her six seconds before impact. However, emergency braking had been disabled to avoid erratic stops, and the safety operator was distracted. Protocols were followed but a life was lost.
Of course, such accidents are rarely the result of technical failure alone. They’re failures of judgement for which no one claims full responsibility because each participant sees only their role in a system that they assume will manage the whole. Regardless of all the ingenuity involved in developing driverless cars, no one asked the only question that could have saved a life: “Should we stop?”
Despite all of its prowess, AI lacks the one capability that matters most in leadership: it does not and cannot care. For all its agentic claims, it has no agency. Data can predict what’s likely to happen but can’t determine what matters. AI can decide fast but cannot reflect. Most critically, it can never ask whether to act.
This is the ethical blind spot where human leadership is critical. It’s also the leadership challenge of the AI age.
Every system has humans behind it. It’s humans who decide what data count, what objectives matter, and what trade-offs are acceptable. When an AI system acts, it only expresses human judgement, even if no single individual ever feels accountable for the answer. And therein lies the danger.
When decisions are increasingly mediated by machines, asking whether something is morally (rather than just factually) right becomes critical. When judgement is the sole human override, ethics are no longer a safeguard at the margins of leadership but become their defining core.
To understand what ethical leadership now requires, let me turn to three philosophers who offer us important lenses through which to view the role of leaders in the age of AI.
First, Aristotle thought of ethics as doing the right thing instinctively rather than following rules. His emphasis reminds us that ethics begin with character. Applied today, leadership isn’t about the cultivation of practical wisdom. In an AI world, the human advantage no longer lies in answers or information — but in questions and formation.
Next, the 19th-century English philosopher John Stuart Mill saw ethics as impact. Good intentions are insufficient if outcomes cause harm. In our technological age, AI will optimize whatever it’s told to optimize regardless of whether it’s at anyone’s expense. Mill suggests that leaders must always account for consequences.
Finally, 20th-century French philosopher Emmanuel Levinas insisted that ethics, just like leadership, begin with encounters. For him, philosophy wasn’t the “love of wisdom,” but the “wisdom of love.” His ethical stance reminds us to stay human within a system that encourages distance. From Levinas we recognize that it isn’t whether a process works, but whether the people it affects feel seen and heard.
Proclaiming that character, impact, and human relationships lie at the core of leadership are not abstract concepts. In Tempe, six seconds were enough to show us what happens when speed replaces judgement and responsibility is diffused across a system. These three lenses help us see the gaps in the age of AI that only leadership can fill. AI can never be attuned to values, impact, and people.
Below are five actions that leaders must take now if they want to lead ethically in an AI-saturated world.
1. Press pause before you press fast forward.There’s something intoxicating about AI’s ability to provide instant answers — especially when our organizations march at a fast beat. Yet no one can reflect quickly. We must slow down to understand the moral tensions, context, and human costs of our decisions.
2. Ask “should we” more often than “can we?”Technical feasibility has never been a moral justification. It’s our role to question the momentum, even and especially when speed is of the essence. Going fast in the wrong direction is never a good idea.
3. Always consider impact as well as intent.An unethical decision doesn’t become ethical just because it was made with the best of intentions. Leaders must ask who benefits, who bears the cost, and who is excluded from the equation.
4. Stay present to people, not just processes.Leadership is social and relational. AI is neither. Decisions affect lives, not dashboards. If no one feels seen in the outcome, something essential has been lost.
5. Know when to stop the system.The ultimate test of leadership is neither speed nor optimization but interruption. Ethics require the courage to slow down, pause, or halt action when responsibility demands it.
I have no doubt that AI will continue to accelerate. As that happens, any ethical lapse, however small, will multiply at scale and at speed. In such a world, ethics can no longer be seen as some kind of philosophical luxury or compliance exercise. They’ve become your competitive advantage.
The real question for leaders is no longer whether their systems work, but whether they have the courage, in the face of relentless pressure to act, to know why and when to stop them.
* * *
Emmanuel Gobillot is among the world’s foremost thinkers and authorities on leadership. Described as “the first leadership guru for the digital generation” and “the freshest voice in leadership today,” he provides consulting to CEOs across countries and industries. A sought-after speaker, he has authored 10 UK and US bestselling books. His new book is Alive Inside: Unlock Your Leadership Advantage in the Age of AI (Routledge, Jan. 22, 2026). Learn more at emmanuelgobillot.com.
—
