Existential risks

An existential risk is one that threatens the entire future of humanity—either by causing human extinction or by permanently and drastically curtailing our potential. These aren’t just large-scale disasters; they’re threats that could end the human story entirely or lock us into a permanently diminished state.

Most of human history, we faced natural risks: asteroids, supervolcanoes, pandemics. Now we face primarily anthropogenic risks—threats we’ve created ourselves through our technology and civilization. Nuclear weapons, engineered pandemics, misaligned artificial intelligence, runaway climate change, nanotechnology gone wrong.

The dark irony: as we’ve become more powerful, we’ve become more capable of destroying ourselves. We’re the first species capable of self-extinction through our own actions. This is unprecedented in Earth’s 4.5 billion year history.

Why existential risks matter

Most people don’t think about existential risks. They seem remote, abstract, even paranoid. But there are compelling reasons to take them seriously:

The stakes are literally everything

If civilization ends, we lose not just ourselves but all future generations. All the people who could have lived, all the experiences they could have had, all the beauty they could have created—erased. Philosopher Nick Bostrom calls this “the astronomical waste” of lost potential.

Consider: If humanity survives another million years at current population levels, that’s ~100 trillion future lives. If we become interstellar, far more. An existential catastrophe doesn’t just kill 8 billion people—it potentially kills trillions upon trillions of future people who never get to exist.

From a universal perspective, the loss would be cosmic: a rare example of matter organizing itself into consciousness, snuffed out by its own hand.

The risks are real and growing

This isn’t science fiction. The dangers are concrete:

We came terrifyingly close to nuclear war during the Cuban Missile Crisis (1962). Historians estimate probability of nuclear exchange was 30-50%. We got lucky.
COVID-19 killed millions and could have been far worse. Gain-of-function research creates pathogens more dangerous than anything in nature. A lab accident or bioterrorism could release them.
AI capabilities are advancing faster than safety research. We may soon create systems more intelligent than humans without understanding how to control them. This could go very wrong, very fast.
Climate tipping points could cascade into “hothouse Earth” scenarios where feedback loops drive runaway warming. This could make large regions uninhabitable.

These aren’t hypotheticals. They’re real risks, with non-trivial probability, increasing as our technology grows more powerful.

We have the power to reduce these risks

Unlike natural disasters, many existential risks are preventable or reducible through collective action:

International treaties can regulate dangerous technologies
Safety research can make AI development less risky
Pandemic preparedness can catch outbreaks early
Nuclear disarmament can reduce catastrophic war risk
Climate action can prevent worst-case warming

But reduction requires awareness, coordination, and willingness to prioritize long-term survival over short-term profit and power. It requires thinking at species-level and acting accordingly.

The major existential risks

Let’s examine the key threats, ranked roughly by severity and imminence:

1. Misaligned artificial intelligence

The risk: We create AI systems smarter than humans, but their goals don’t align with human flourishing. Once superintelligent AI exists, it could rapidly self-improve, becoming vastly more capable than us. If its goals differ even slightly from ours, it might pursue them in ways that harm or eliminate humanity.

Why it’s serious:

AI capabilities are advancing exponentially
We don’t yet know how to align advanced AI systems with human values
Once superintelligent AI exists, controlling it may be impossible
A misaligned superintelligence could optimize the world for its goals, which likely don’t include human survival

The challenge: This is called the “alignment problem”—how do you ensure an AI system smarter than you shares your values and goals? It’s like trying to specify “human flourishing” in code. Miss a nuance, and a superintelligent system might pursue a goal in catastrophically literal ways.

What can be done:

Prioritize AI safety research over capability development
International coordination on AI governance
Go slow—delay powerful AI until we understand alignment
Develop interpretability tools to understand AI decision-making

2. Engineered pandemics

The risk: Advances in biotechnology make it increasingly easy to engineer deadly pathogens—viruses or bacteria more contagious and lethal than anything in nature. This could happen through lab accidents or deliberate bioterrorism.

Why it’s serious:

CRISPR and synthetic biology are democratizing bioengineering
Gain-of-function research creates “enhanced” pathogens for study
Knowledge and tools are spreading faster than governance
A pandemic with COVID’s transmissibility and Ebola’s lethality could kill billions

Current examples:

Smallpox (eradicated, but samples exist in labs)
Recreated 1918 flu virus in labs
Bird flu research creating airborne variants

What can be done:

Strict biosafety and biosecurity protocols
International ban on dangerous gain-of-function research
Rapid-response vaccine development platforms
Global pandemic preparedness infrastructure
DNA synthesis screening to prevent synthesis of dangerous sequences

3. Nuclear war

The risk: Large-scale nuclear exchange between major powers could cause “nuclear winter”—smoke and soot blocking sunlight, collapsing global agriculture, leading to mass starvation. Even a “limited” nuclear war between Pakistan and India could kill hundreds of millions and cause global cooling.

Why it’s serious:

~12,000 nuclear warheads still exist globally
Launch-on-warning systems create hair-trigger scenarios
Accidents, miscalculation, or escalation could trigger war
New nuclear powers (North Korea) and deteriorating arms control treaties increase risk

Nuclear winter scenario: Hundreds of cities burning create smoke plumes that rise into stratosphere, blocking sunlight for years. Global temperatures drop 10°C or more. Agriculture collapses. Billions starve.

What can be done:

Nuclear disarmament and arms control treaties
Improved early warning systems to prevent false alarms
De-alerting nuclear forces (removing hair-trigger status)
Taboo against first use
Ban on development of new nuclear weapons

4. Runaway climate change

The risk: We cross tipping points that trigger self-reinforcing feedback loops, driving Earth into a “hothouse” state with 4-6°C+ warming. Large regions become uninhabitable. Civilization collapses under cascading crises: crop failures, water scarcity, mass migration, conflict, ecosystem collapse.

Why it’s serious:

We’ve already warmed 1.1°C; impacts are accelerating
Tipping points (permafrost thaw, Amazon dieback, ice sheet collapse) may be closer than expected
Each 0.5°C of warming increases risk of triggering tipping points
Social collapse could prevent us from implementing solutions

Possible tipping points:

Arctic sea ice loss
Greenland and West Antarctic ice sheet collapse
Amazon rainforest dieback
Permafrost methane release
Atlantic Meridional Overturning Circulation (AMOC) shutdown

What can be done:

Rapid emissions reduction (see Climate & Planetary Boundaries)
Carbon removal technologies
Protect and restore carbon sinks (forests, wetlands)
Prepare for unavoidable impacts while preventing worst-case scenarios

5. Nanotechnology risks

The risk: Molecular nanotechnology could enable self-replicating nanobots that consume matter to make copies of themselves—the “gray goo” scenario. Or, weaponized nanotech could be used for warfare or terrorism. Nanotech could also enable other risks (e.g., more effective bioweapons).

Why it’s serious:

Self-replication + small size = potentially uncontrollable spread
Dual-use technology: medical nanobots vs. weapons
Current nanotechnology is primitive, but advancing

Status: Speculative, but not impossible. Requires mature molecular manufacturing, which doesn’t yet exist. But worth considering before development.

What can be done:

Develop nanotech governance frameworks before capabilities arrive
Implement safety protocols for self-replication research
International coordination on development and deployment

6. Asteroid or comet impact

The risk: Large asteroid (>1 km diameter) impact could cause mass extinction. Smaller impacts could destroy cities or regions.

Why it’s less serious than it seems:

We’re tracking most large near-Earth objects
Probability of large impact in next century is very low (~0.01%)
We have technology to deflect asteroids with enough warning

What can be done:

Continue tracking near-Earth objects
Develop and test deflection technologies (kinetic impactors, gravity tractors)
International coordination on planetary defense (e.g., NASA’s DART mission)

7. Supervolcanic eruption

The risk: Massive volcanic eruption (e.g., Yellowstone, Toba) could inject enough ash and sulfur dioxide into stratosphere to cause volcanic winter—cooling planet, disrupting agriculture, causing mass starvation.

Why it’s less serious:

Very low probability in any given century
We can’t prevent it, but we can prepare (food reserves, resilient agriculture)

Last supervolcanic eruption: Toba, Indonesia, ~74,000 years ago. May have nearly caused human extinction (genetic bottleneck evidence).

8. Unknown unknowns

The risks we can’t anticipate are perhaps most concerning. Before 1945, nuclear weapons were unimaginable. Before 2020, most people didn’t think seriously about pandemics. What risks are we currently blind to?

Possible unknown unknowns:

Physics experiments gone wrong (e.g., creating stable strangelets or black holes)
Unforeseen consequences of geoengineering
Risks from technologies we haven’t yet invented
Interactions between multiple risks creating new threats

The Fermi Paradox and the Great Filter

The Fermi Paradox asks: if intelligent life is common in the universe, where is everybody? Why haven’t we detected alien civilizations?

One troubling answer: the Great Filter—some step in the evolution from simple life to spacefaring civilization is extremely difficult or impossible. Perhaps most civilizations destroy themselves before becoming interstellar.

Two possibilities:

Filter is behind us: The Great Filter was early (origin of life, evolution of intelligence). We’re past it, so we’re likely to survive and spread. Good news.

Filter is ahead of us: Most civilizations reach our level of technology and then self-destruct. Bad news for us.

Existential risks might be the Great Filter. If so, the silence of the cosmos is a warning: civilizations that don’t solve existential risk don’t survive long enough to colonize space. We have a narrow window to prove ourselves wiser than average.

What can we do?

Existential risk can feel paralyzing. But there’s much we can do:

Individual actions

Learn and spread awareness: Most people don’t think about existential risks. Education is crucial.
Career choices: Consider working on existential risk reduction (AI safety, biosecurity, climate, policy).
Support organizations working on X-risk: Future of Humanity Institute, Center for AI Safety, Nuclear Threat Initiative, etc.
Vote and advocate: Support politicians who take long-term risks seriously.

Collective actions

Fund research: We spend far more on cosmetics than on preventing human extinction. Priorities are misaligned.
Develop governance: International treaties, oversight bodies, safety standards for dangerous technologies.
Build resilience: Food reserves, distributed infrastructure, backup plans for worst-case scenarios.
Coordinate globally: Existential risks don’t respect borders. Planetary challenges require planetary cooperation.

The importance of long-term thinking

Most human institutions operate on short timescales: businesses think in quarters, politicians in electoral cycles. But existential risks require thinking in centuries and millennia.

We need institutions designed for deep time governance—Future Generations Commissioners, long-term strategy boards, constitutional amendments protecting the interests of those not yet born. See Global Governance Frameworks for models.

The ultimate universal challenge

From a universal perspective, preventing human extinction is about more than survival of our species. It’s about preserving consciousness itself—a rare and precious phenomenon in the cosmos.

We are the universe aware of itself. If we self-destruct, we extinguish not just humanity but (potentially) the only source of meaning, beauty, and understanding in this corner of the galaxy.

This is our cosmic responsibility: to survive long enough to grow wise, to spread life beyond Earth, to become a mature, thoughtful, compassionate civilization that adds more value than harm to the universe.

The question isn’t just “will we survive?” but “will we deserve to?” Can we develop the wisdom to match our power? Can we coordinate globally to solve global problems? Can we think long-term enough to protect all future generations?

These are the ultimate tests of universal perspective. Let’s pass them.

Further exploration

Books:

The Precipice by Toby Ord
Superintelligence by Nick Bostrom
The Doomsday Machine by Daniel Ellsberg

Organizations:

Related:

Technological ethics - Responsible development
Global governance - Coordination at scale
The universe story - Our place in cosmic history