Yes, AI could contribute to a catastrophe, but no reliable evidence shows it is destined to destroy humanity—or that it is already the biggest threat. People are responsible for current risks such as nuclear escalation, climate disruption and biological threats. AI could amplify some of those dangers through misuse or military deployment; a future loss of control by highly capable systems is a separate, uncertain possibility. Which threat is “bigger” depends on the scenario and time horizon, and the available estimates do not provide a fair, like-for-like ranking.
What does “AI destroying us all” mean?
The phrase can refer to very different outcomes: a system causing deaths or widespread disruption, a catastrophe that severely damages civilization, or human extinction. Those outcomes should not be treated as interchangeable. Nor does every use of “existential risk” mean the same thing: the Bulletin of the Atomic Scientists, for example, uses the term when discussing a conditional danger from putting AI in control of critical physical systems, while surveys may ask respondents specifically about human extinction.
As an Amazon Associate I earn from qualifying purchases.
It also matters whether AI is the direct source of harm or a tool that helps people cause it. A machine acting beyond effective human oversight is not the same scenario as a person using AI to spread deception or assist a dangerous activity.
How could AI contribute to catastrophe?
People using AI to magnify harm
AI can assist or scale harmful human activity. The Bulletin’s 2024 Doomsday Clock statement discusses AI-enabled disinformation, biological misuse and military applications, including lethal autonomous weapons. Disinformation can also make it harder for people and institutions to respond to other threats. These are concerns about how people develop, deploy or exploit tools; they do not establish that AI has independently caused a catastrophe.
#1 Best Overall
Military integration and escalation
AI could affect the speed or stability of military decisions, including if it were integrated into systems related to nuclear weapons. The Bulletin warned that decisions to put AI in control of important physical systems—“in particular, nuclear weapons”—could pose a direct existential threat. That is a warning about a conditional deployment choice, not evidence that AI controls nuclear launch decisions today.
The underlying danger of nuclear escalation does not depend on AI: human decisions and miscalculation are already pathways to catastrophe. The World Economic Forum’s 2024 Global Risks Report also discussed potential risks from AI in nuclear systems. Any added risk would depend on how such systems were designed and used.
Rank #2
Future loss of control
A different concern is that a future, highly capable AI system might act in ways people cannot effectively oversee or stop. This scenario depends on assumptions about future capabilities, how systems behave, and whether safeguards work. The Associated Press’s 2026 reporting on the debate around advanced models and human control says there is no agreed likelihood or timing; it also describes the 2026 International AI Safety Report’s account of the risk as unusually ambiguous. This is a debated future possibility, not proof that current systems can escape human control or that superintelligence is inevitable.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Which dangers already exist without AI?
AI is not required for several major catastrophic pathways. Nuclear conflict can arise from human decisions and miscalculation. Climate disruption is driven by human-caused warming, and the World Economic Forum’s 2024 report discussed potential tipping risks and systemic effects on food, water and health. Climate catastrophe does not automatically mean human extinction: the particular harm and time horizon matter.
Biological threats, including pandemics and biological misuse, are another concern independent of AI. AI might affect how people pursue harmful activities, but the sources considered here do not establish a numerical probability that AI-enabled biological misuse would cause human extinction. RAND’s 2025 scenario analysis considers possible routes involving pathogens, nuclear weapons and geoengineering. It concludes that creating an AI-driven human-extinction threat would be immensely challenging under the scenarios it examines, while explicitly not ruling it out.
What do the numbers say—and what don’t they say?
The figures below measure different things. Expert forecasts are conditional judgments about possible futures; public survey results measure what respondents believe. Neither is an observed rate of catastrophe or a definitive scientific probability.
| Figure | What it measures | How to interpret it |
|---|---|---|
| 10% | Median expert forecast in the Longitudinal Expert AI Panel (LEAP) Wave 9 for AI-caused catastrophe by 2100, conditional on rapid AI capability progress. | A conditional forecast, not an observed frequency or consensus probability. |
| 15% | LEAP Wave 9 median forecast for catastrophe from any cause by 2100, under the same rapid-progress condition. | It shares the rapid-progress condition with the 10% estimate but covers catastrophe from any cause, not AI alone. |
| 67% | LEAP Wave 9 median share of total global catastrophic risk attributed to AI by 2100 in a world of rapid AI capability progress. | A conditional expert judgment about AI’s share, not the probability of an AI catastrophe. |
| About 30% and 53% | LEAP Wave 9’s approximate shares of total catastrophic risk attributed to AI by 2100 under slow and moderate AI progress, respectively. | Conditional expert forecasts; the shares vary with the assumed pace of progress. |
| 13%, 42% and 21% | In the 2024 SARA technical report, the shares of surveyed Australian respondents who selected AI, nuclear war and climate change, respectively, as the most likely cause of human extinction among six options. | Public perceptions in a forced-choice survey, not a ranking of actual extinction probabilities. |
| 90 seconds to midnight | The Bulletin’s symbolic Doomsday Clock setting in its 2024 statement. | A warning indicator, not a calibrated probability or countdown. |
LEAP’s forecasts also show why a single percentage can mislead: its panel’s estimate of AI’s share changes with assumptions about how quickly capabilities advance. Respondents disagree about how much AI adds to overall risk; some see it as entangled with many catastrophic pathways, while others emphasize threats such as pandemics, world war and climate change that exist independently of AI.
So are humans still the bigger threat?
For risks already unfolding, people remain the direct agents: they make decisions about nuclear weapons, drive climate change, and can misuse biological tools. AI may amplify some human actions or introduce additional risks if future systems become more capable and harder to control. That makes “humans versus AI” a false divide in many scenarios: human choices determine whether and how AI becomes part of a danger.
Best Value
There is no comprehensive estimate in the cited sources that compares AI, nuclear war, pandemics and climate change using the same time horizon and definition of catastrophe. Without that common basis, declaring one the numerically bigger threat would overstate what the evidence establishes. A more useful comparison asks who or what drives the pathway, what is already happening versus assumed about the future, how severe and reversible the harm could be, and where intervention is possible.
What can reduce the risk?
The sources point toward governance and deployment choices rather than a prediction that catastrophe is inevitable. The Bulletin argues for broader AI governance. The United Nations High-Level Advisory Body on AI’s final report, Governing AI for Humanity, released in September 2024, frames international cooperation as necessary and addresses gaps in current arrangements.
- Keep meaningful human responsibility and oversight over consequential physical and military systems.
- Assess how AI could enable misuse or amplify existing threats, including disinformation and biological risks.
- Coordinate governance internationally, since risks and responses can cross borders.
- Judge claims against their assumptions: distinguish a forecast from an observed outcome, and a conditional warning from evidence that the scenario is happening now.
The case for taking AI risk seriously does not require treating it as the only danger—or assuming a worst-case future is certain. Humans already shape major catastrophic risks, and human decisions will influence whether AI reduces, leaves unchanged or increases them.
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.
