AI Existential Risk Without Panic: How Leaders Should Communicate Serious Uncertainty

When leaders use the phrase “existential risk,” they face a psychological problem as well as a technical one. Speak too softly and people may ignore a serious danger. Speak too dramatically and the warning can produce panic, denial or helplessness.

On 7 September 2026, UN High Commissioner for Human Rights Volker Türk told the Human Rights Council that advanced artificial intelligence could pose an existential risk to humanity and called for an all-out effort to establish strong safety and security guarantees. He also warned about concentrated power and called for agreed red lines among countries hosting AI development and its supply chains.

The warning is important, but it is not a probability estimate and it does not prove catastrophe is imminent. It is a policy demand made under uncertainty. Responsible communication must keep those categories separate.

Why extreme language can backfire

Human attention is sensitive to vivid threats. A dramatic scenario can make a low-frequency risk feel immediate because it is easy to imagine. This is related to the availability heuristic: the mind often judges likelihood partly by how readily examples come to mind.

The opposite response also occurs. When warnings are repeated without concrete actions, people develop alarm fatigue. Some protect themselves emotionally by dismissing the message; others accept the danger but conclude that nothing they do matters. Both reactions reduce constructive engagement.

Effective risk communication increases agency. It should help people understand what could happen, what remains uncertain and which safeguards can reduce the danger.

Five questions every warning should answer

1. What is the hazard? Name the capability or failure mode rather than using “AI” as one undifferentiated threat.
2. What is the pathway? Explain how the failure could move from a model behaviour to real-world harm.
3. What is the evidence? Separate observed incidents, expert inference, simulations and speculation.
4. Who is exposed? Identify the people, institutions or systems that would bear the risk.
5. What can be done now? Pair the warning with controls, responsibility and a timetable.

Existential risk is not the only human-rights issue

A focus on humanity-ending scenarios can overshadow harms already experienced: discrimination, surveillance, labour displacement, manipulation, fraud and autonomous weapons. Türk’s speech connected AI safety with democratic institutions, critical infrastructure and concentrated private power. That wider framing matters because governance cannot wait for a hypothetical final catastrophe.

Near-term and extreme risks are not necessarily competitors. Strong identity controls, cybersecurity, evaluation, accountability and human command can reduce current harm while also building institutions capable of responding to more powerful systems.

How leaders can communicate uncertainty honestly

• Use ranges and scenarios instead of presenting one forecast as certainty.
• State what evidence would cause the assessment to change.
• Avoid anthropomorphic language unless it describes observed behaviour precisely.
• Distinguish a system’s capability from its intention or consciousness.
• Name the organisation responsible for each mitigation.
• Report incidents and near misses in a form independent experts can evaluate.
• Explain the cost of both action and inaction.

This approach protects credibility. If a leader exaggerates today, audiences may discount a more urgent warning tomorrow. If a leader offers empty reassurance, a later failure can destroy trust in the entire institution.

The psychology of concentrated power

Türk also warned that a small number of people hold extraordinary influence over AI. Concentration can create a diffusion-of-responsibility paradox. The public assumes powerful firms must have control; companies assume governments will define the rules; governments wait for technical certainty. Everyone recognises the risk while responsibility remains psychologically elsewhere.

Clear governance reverses that pattern. It assigns duties before an incident: who evaluates, who authorises deployment, who receives protected whistleblowing reports, who can order a pause and who compensates those harmed.

From fear to calibrated trust

MaryChuks has argued that trust should be calibrated rather than absolute. Read Why We Trust AI, strengthen the AI verification habit, and compare this warning with OpenAI’s stronger guardrail threshold.

Use Insight AI 360 to examine claims, competing explanations and evidence before accepting a confident conclusion.

Discussion question: Which creates more danger in public AI debate—exaggerated fear, or reassurance that hides genuine uncertainty?

Source

• Reuters: AI could pose existential risk to humanity, UN rights chief warns, 7 September 2026