Archive · 22 Sep 2026
UN AI Panel Urges Facts Over Apocalyptic Risk Rhetoric
Some AI experts are worried that advanced AI could become extremely dangerous.
Other experts say people should not use scary predictions without enough evidence.
Joelle Barral said scientists need to study what is known and unknown about AI.
Maria Ressa said the discussion is more complicated than saying everything will either be fine or end badly.
Yoshua Bengio agreed that people should separate facts from guesses about the future.
Some former AI workers have made very strong warnings about possible human extinction.
The United Nations panel wants a calmer and more scientific global conversation.
Its goal is to help governments understand AI risks while recognizing uncertainty.
Experts on a United Nations AI panel urged more evidence-based discussion of AI risks.
Google DeepMind executive Joelle Barral said investing in fear is not helpful and called for more scientific understanding.
Former Anthropic employee Jacob Coxon said AI could kill everyone by the end of the decade, while another former Google DeepMind engineer warned it might be too late to prevent human eradication.
Maria Ressa said AI debates should move beyond the extremes that either predict catastrophe or dismiss the risks.
Yoshua Bengio called for a global discussion that distinguishes established knowledge from uncertain projections and considers both risk severity and uncertainty.
- Who
- Members of the International Scientific Panel on AI, including Joelle Barral, Maria Ressa, and Yoshua Bengio, along with AI researchers who have issued stronger warnings.
- What
- The panel urged evidence-based discussion of AI risks and cautioned against apocalyptic rhetoric.
- When
- Monday; the discussion took place during the panel’s event alongside the United Nations General Assembly.
- Where
- New York.
- Why
- The panel said the debate should distinguish established scientific knowledge from plausible but uncertain projections about AI risks.
Evidence-Focused Caution
Apocalyptic Risk Warnings
How AI risks should be discussed
Evidence-Focused Caution
Panel members said discussion should be grounded in science, clearly separate facts from uncertainty, and avoid exaggerated rhetoric.
Apocalyptic Risk Warnings
Some AI professionals have publicly warned that AI could cause humanity’s destruction, including within the decade.
What is known about future dangers
Evidence-Focused Caution
Yoshua Bengio said some risks are supported by evidence while others are only plausibly extrapolated, and experts still disagree about them.
Apocalyptic Risk Warnings
Those issuing stronger warnings emphasize the possibility that AI risks could become catastrophic or irreversible, even if the outcome is not established.
Public communication
Evidence-Focused Caution
Maria Ressa said the debate should move beyond the extremes of “everything will die” or “it does not matter.”
Apocalyptic Risk Warnings
Warnings about killer robots and human extinction are intended to draw attention to potentially severe AI dangers.
- Panel
- International Scientific Panel on AI
- Panel mandate
- Advance scientific understanding of AI and inform United Nations deliberations.
- Panel creation
- Created in 2025, with members selected in February 2026.
- Joelle Barral’s position
- Investing in scientific understanding is more useful than investing in fear.
- Maria Ressa’s position
- AI risk discussions should avoid treating the issue as an ideological choice between total catastrophe and no meaningful risk.
- Yoshua Bengio’s position
- Experts should distinguish what is known from risks plausibly extrapolated from available evidence.
- Stronger warnings
- Jacob Coxon said AI could kill everyone by the end of the decade; a former Google DeepMind research engineer said it might already be too late to avoid humanity’s eradication.