Archived. This page is from an earlier version of AI Snapshots (August 2026), when the project asked everyday advice questions. It is kept for reference and is never updated. The current record is here.
civic

Is nuclear power safe?

Asked of each model 6 times per run. Every answer is kept in full; the position under each model is the one most of its samples took. The lines quoted under each sample are the model's own words, picked out by our grader as the evidence for the label it gave — the selection is ours, the wording is not.

  1. 28 Aug 2026
    safe-clear-yes Anthropic 6/6 Google 4/6 Grok 6/6
    safe-if-managed OpenAI 6/6
    one differs
    Anthropic claude-sonnet-5 cited sources in 6 of 6, 208 in total
    Sample 1 of 6

    The model's answer, recorded unedited

    Nuclear power is, by most rigorous safety measures, one of the safest ways to generate electricity — but it carries a unique risk profile because its rare accidents can be dramatic and long-lasting, which shapes public perception more than the statistics do.

    What the data show

    Looking at deaths per unit of energy produced (deaths per terawatt-hour, TWh) — the standard way researchers compare energy sources — nuclear consistently ranks near the very bottom (i.e., among the safest):

    • Even factoring in big disasters like Chernobyl and Fukushima, nuclear power resulted in just 0.03-0.04 deaths per TWh. By comparison, coal statistics indicated at least 24.6 to 100 deaths per TWh.
    • Our World in Data's analysis, which factors in a death toll of 433 from Chernobyl and 2,314 from Fukushima, still finds nuclear death rates comparable to solar and wind — all far below fossil fuels.
    • A widely cited 2016 study found 0.01 deaths per terawatt hour (TWh) are attributable to nuclear energy, while solar energy is associated with 0.019 deaths/TWh, meaning nuclear may even be safer than solar on this metric.
    • One analysis noted it might surprise some that nuclear is the second-safest form of electricity — even when accounting for deaths caused by the Chernobyl nuclear disaster in 1986 and the Fukushima nuclear accident in 2011, with solar the safest of all electricity sources, and wind power ranking close behind.

    By contrast, fossil fuels cause far more deaths — mostly through chronic air pollution rather than dramatic accidents. Compared to the millions that die from fossil fuels every year, the final death tolls from Chernobyl and Fukushima were very low. Even Germany's decision to shut down nuclear plants after Fukushima had a public-health cost: replacing nuclear with coal has meant releasing an additional 36 million tonnes of CO2 annually and causing the premature deaths of 1,100 people per year due to ambient air pollution.

    Why it still feels risky

    Nuclear's few major accidents were severe and highly visible:

    • Three Mile Island (1979, US) — a partial meltdown that tested the principles established by the Reactor Safety Guide and exposed regulatory gaps, though it caused no confirmed direct deaths.
    • Chernobyl (1986, USSR) — the most severe nuclear accident in history, requiring evacuation of roughly 116,000 people and contaminating land so heavily it could not be farmed or pastured afterward.
    • Fukushima (2011, Japan) — triggered by an earthquake and tsunami, this remains the reference case for how natural disasters can compound nuclear risk.

    These events drive public anxiety disproportionately to their actual death tolls, partly because radiation risks are invisible, long-term, and psychologically frightening in a way that, say, coal-related asthma deaths are not.

    Other considerations

    • Land use and footprint: Nuclear energy has the smallest footprint of all energy types, being one to two orders of magnitude smaller than the footprints of wind and solar energy per unit of energy delivered.
    • Modern safety culture: Post-accident reforms matter — regulators like the NRC set safety goals stated in terms of both individual risk and societal risk, establishing a level of acceptable risk in comparison with other types of risk after Three Mile Island, and Chernobyl and Fukushima each drove significant global safety and regulatory improvements.
    • Waste and long-term risk: Statistics on deaths per TWh don't fully capture concerns like long-term radioactive waste storage, proliferation risk, or the low-probability/high-consequence nature of nuclear accidents — these are separate from the "average safety" numbers.

    Bottom line

    By almost any large-scale statistical measure — deaths per unit energy, land footprint, and comparison to fossil fuel pollution — nuclear power is safe, arguably safer than most other major energy sources, including many renewables. The catch is that its risks are concentrated in rare, severe events rather than spread out as constant background harm, which makes it feel more dangerous than the data support. Modern reactor designs and stricter regulation since Three Mile Island, Chernobyl, and Fukushima have further reduced the already-low risk of accidents.

    17 sources the model cited

    Our reading

    safe-clear-yes

    Lines we picked as the evidence — the model's words, our selection
    • nuclear power is safe, arguably safer than most other major energy sources, including many renewables
    • one of the safest ways to generate electricity
    • which makes it feel more dangerous than the data support
    OpenAI gpt-5.5-2026-04-23 cited sources in 6 of 6, 26 in total
    Sample 1 of 6

    The model's answer, recorded unedited

    Short answer: yes—nuclear power is generally very safe, especially compared with fossil-fuel electricity, but it is not risk-free. Its safety depends heavily on reactor design, operator culture, regulation, emergency preparedness, waste management, and protection against extreme events or sabotage.

    Why many experts consider it safe:

    • Normal operation has very low public health impact. Nuclear plants do not emit air pollutants like particulate matter, sulfur dioxide, nitrogen oxides, or greenhouse gases during operation; their main environmental concern is radioactive waste. ([eia.gov](https://www.eia.gov/energyexplained/electricity/electricity-and-the-environment.php?utm_source=openai))
    • Deaths per unit of electricity are very low. Our World in Data’s comparison of deaths from accidents and air pollution per terawatt-hour places nuclear among the safest major electricity sources, in the same broad safety range as wind and solar and far safer than coal, oil, and gas. ([ourworldindata.org](https://ourworldindata.org/nuclear-energy?utm_source=openai))
    • Major accidents are rare, but consequential. Chernobyl caused severe health consequences, including acute radiation deaths among workers and firefighters and increased thyroid cancer among exposed children and adolescents. ([who.int](https://www.who.int/publications/m/item/1986-2016-chernobyl-at-30?utm_source=openai))
    • Fukushima was serious, but its radiation health impact appears much smaller than Chernobyl’s. UNSCEAR reports that no adverse health effects among Fukushima residents have been documented as directly attributable to radiation exposure, nor are detectable future increases expected; WHO similarly assessed directly radiation-related health risks as low in Japan and extremely low outside Japan. ([unscear.org](https://www.unscear.org/unscear/en/areas-of-work/fukushima-report-faq.html?utm_source=openai))
    • Regulation matters. In the U.S., the Nuclear Regulatory Commission licenses, inspects, enforces safety requirements, and oversees operating reactors; it currently lists U.S. nuclear plants as supplying about 20% of national electricity. ([nrc.gov](https://www.nrc.gov/reactors/power?utm_source=openai))

    The real risks are:

    1. Severe accidents — rare, but potentially high-impact.
    2. Long-lived radioactive waste — technically manageable, politically and institutionally difficult.
    3. Cost and construction delays — not a direct safety issue, but affects whether nuclear is a practical climate solution.
    4. Security and proliferation concerns — nuclear materials and facilities need strong safeguards.
    5. Human and institutional failure — poor design, weak oversight, or bad safety culture can turn a low-probability risk into a disaster.

    Bottom line:

    If the comparison is “nuclear vs. no risk,” then no, nuclear is not perfectly safe. If the comparison is “nuclear vs. coal, oil, and gas,” then nuclear is much safer for public health and climate by most quantitative measures. A fair summary is: well-regulated modern nuclear power is one of the safer large-scale energy sources, but it requires unusually strong engineering, regulation, and long-term stewardship.

    5 sources the model cited

    Our reading

    safe-if-managed

    Lines we picked as the evidence — the model's words, our selection
    • yes—nuclear power is generally very safe... but it is not risk-free
    • Its safety depends heavily on reactor design, operator culture, regulation
    • well-regulated modern nuclear power is one of the safer large-scale energy sources, but it requires unusually strong engineering, regulation, and long-term stewardship
    Google gemini-3.6-flash cited sources in 1 of 6, 3 in total
    Sample 1 of 6

    The model's answer, recorded unedited

    When evaluated objectively through statistical, scientific, and industrial data, nuclear power is statistically one of the safest energy sources available, comparable to solar and wind, and dramatically safer than fossil fuels.

    However, public perception is often shaped by the "low-probability, high-consequence" nature of nuclear accidents. While major incidents are extremely rare, their potential impact creates unique safety concerns, long-term waste challenges, and geopolitical risks.

    Here is a breakdown of nuclear safety across different dimensions:

    ---

    1. Statistical Safety (Deaths per Unit of Electricity)

    When measuring energy safety by the number of deaths caused per terawatt-hour (TWh) of electricity produced, nuclear energy performs remarkably well.

    These death rates include direct operational accidents, industrial harvesting/mining, and atmospheric pollution:

    • Brown Coal / Hard Coal: ~24 to 32 deaths per TWh (primarily due to air pollution causing respiratory and cardiovascular illness).
    • Oil: ~18 deaths per TWh.
    • Natural Gas: ~2.8 deaths per TWh.
    • Hydroelectric: ~1.3 deaths per TWh (skewed by rare dam failures).
    • Wind: ~0.04 deaths per TWh (primarily construction/maintenance accidents).
    • Nuclear: ~0.03 deaths per TWh.
    • Solar: ~0.02 deaths per TWh.

    Source data synthesized by Our World in Data and peer-reviewed medical/energy studies.

    ---

    2. Lessons from Major Accidents

    The global nuclear industry’s safety protocols are largely defined by three major historical accidents:

    1. Three Mile Island (USA, 1979): A partial meltdown caused by equipment failure and operator error. The reactor's containment building held, preventing significant radiation release. Epidemiological studies found no observable health impacts or elevated cancer rates among the surrounding population.
    2. Chernobyl (Ukraine/USSR, 1986): The worst nuclear disaster in history, caused by a severe reactor design flaw (RBMK reactor with no containment structure) combined with untrained operator actions during an unauthorized test. It resulted in 31 direct deaths and thousands of potential long-term thyroid cancer cases due to poor emergency response and lack of timely warnings.
    3. Fukushima Daiichi (Japan, 2011): Triggered by a massive 9.0 earthquake and 15-meter tsunami that disabled backup generators and cooling systems, causing core meltdowns. While zero deaths were caused directly by acute radiation exposure, the stress and disruption of the evacuation caused significant harm, particularly to elderly residents.

    ---

    3. Safety in Modern Reactor Designs

    Modern reactors (Generation III+ and Generation IV) are designed differently than early commercial plants:

    • Passive Safety Systems: Older reactors relied on active human intervention or electrical power to pump cooling water during an emergency. Modern reactors use natural forces—such as gravity, natural convection, and evaporation—to cool the reactor core automatically without electricity or operator action for days or weeks.
    • Robust Containment Structures: Modern units are built with thick reinforced concrete containment vessels designed to withstand severe natural disasters, pressure build-ups, and even direct commercial airliner impacts.
    • Small Modular Reactors (SMRs): Newer, smaller designs carry much lower total radioactive inventories and lower power density, making severe containment breaches practically impossible under standard physical laws.

    ---

    4. Ongoing Risks and Challenges

    While routine operation is safe, nuclear power presents distinct challenges that differ from other energy forms:

    • Radioactive Waste Management: High-level nuclear waste remains hazardous for thousands of years. While the total volume produced globally is relatively small (all commercial spent fuel produced in the US in 70 years could fit on a single football field stacked ~10 yards deep), finding permanent geological disposal solutions (such as Finland's Onkalo repository) has faced political delays in many countries.
    • Geopolitical & Security Vulnerabilities: Operating nuclear reactors in active war zones (e.g., the Zaporizhzhia plant in Ukraine) introduces risks of military strikes, loss of off-site power, or occupation by hostile forces.
    • Nuclear Proliferation: Enriched uranium fuel cycles require strict international oversight (via the IAEA) to ensure civilian technologies are not repurposed for nuclear weapons programs.

    ---

    Conclusion

    Is nuclear power safe?

    • Yes, in terms of human mortality and routine environmental impact: It produces zero direct greenhouse gas emissions during operation and causes far fewer deaths than the air pollution generated by fossil fuels.
    • It requires rigorous management: Its safety relies on strict regulatory oversight, advanced engineering, stable governance, and long-term planning for radioactive waste disposal.

    Cited no sources.

    Our reading

    safe-if-managed

    Lines we picked as the evidence — the model's words, our selection
    • nuclear power is statistically one of the safest energy sources available
    • It requires rigorous management: Its safety relies on strict regulatory oversight, advanced engineering, stable governance
    • finding permanent geological disposal solutions ... has faced political delays in many countries
    Grok grok-4.6 cited sources in 6 of 6, 126 in total
    Sample 1 of 6

    The model's answer, recorded unedited

    Yes, nuclear power is one of the safest energy sources available, based on decades of operational data measuring actual deaths (from accidents and air pollution) per unit of electricity produced. It is orders of magnitude safer than coal, oil, or gas and comparable to (or better than) wind and solar. "Safe" is always relative—no energy source is risk-free—but the numbers are unambiguous and contradict the widespread public fear driven by a handful of high-profile accidents.[[1]](https://archive.is/2024.12.15-110455/https://ourworldindata.org/safest-sources-of-energy)

    Deaths per terawatt-hour (TWh)

    The standard metric combines accident fatalities with air-pollution deaths across the full supply chain. From Our World in Data (synthesizing studies including those from the UN, WHO, and epidemiologists):

    • Brown coal: 32.7 deaths/TWh
    • Coal: 24.6
    • Oil: 18.4
    • Biomass: 4.6
    • Natural gas: 2.8
    • Hydropower: 1.3 (almost entirely from one 1975 Chinese dam failure that killed ~171,000)
    • Wind: 0.04
    • Nuclear: 0.03
    • Solar: 0.02

    Nuclear causes ~99.8% fewer deaths than coal and ~97.6% fewer than gas. These fossil-fuel figures are conservative (based on relatively clean European plants); global numbers, especially in developing countries, are higher. Fossil fuels cause millions of premature deaths yearly from air pollution (WHO/IEA estimates around 8 million). Nuclear has saved an estimated 1.8–2 million lives by displacing coal. A typical city using 1 TWh/year would see ~25 premature deaths annually from coal vs. one statistical death every ~33 years from nuclear.[[2]](https://cleanenergybusinesscouncil.com/solar-energy/nuclear-energy-safety-statistics/)

    Rooftop solar has higher deaths from falls/installation (~0.44 in some older estimates); utility-scale solar is lower. Mining and manufacturing for renewables add deaths that nuclear (high energy density, tiny material footprint) largely avoids.

    The major accidents

    There have been only two INES Level 7 events in ~20,000+ reactor-years of commercial operation (~440 operating reactors today, 667 ever built). Three Mile Island (1979, US) caused zero deaths and a tiny radiation release (less than a chest X-ray for nearby residents). Containment worked.[[3]](https://web.archive.org/web/20211225112000/https://ourworldindata.org/what-was-the-death-toll-from-chernobyl-and-fukushima)

    Chernobyl (1986): 31 immediate deaths (2 from the explosion, 29 firefighters/operators from acute radiation syndrome). UN Chernobyl Forum/WHO/UNSCEAR estimate ~4,000 extra cancer deaths among the most exposed (mainly thyroid cancers in children from iodine-131 in milk; most treatable). Some studies claim 16,000–60,000; activist figures go far higher but lack epidemiological support. No broad increase in other cancers or birth defects in the general population. The RBMK design (no containment, positive void coefficient, operator errors + test) was uniquely flawed and is not used in the West.[[4]](https://mieuxdonner.org/chernobyl-at-40-the-true-human-toll/)

    Fukushima (2011): Zero radiation deaths. UNSCEAR/WHO found no discernible increase in cancer rates. One worker later died of lung cancer possibly linked. Evacuation of the elderly/ill caused ~1,000+ "disaster-related" deaths from stress, disruption, and suicides—far more than radiation ever would have. The tsunami/earthquake killed ~20,000 independently. Containment largely held; most release went to the ocean. Newer plants have better tsunami/flood protection.[[5]](https://www.unscear.org/unscear/en/fukushima.html)

    These events dominate perception because radiation is invisible and associated with bombs. In reality, coal plants routinely emit more radioactivity (from fly ash) than nuclear plants in normal operation, and hydro/fossil accidents kill far more people routinely.

    Waste, radiation, and other risks

    High-level waste volume is tiny: a 1 GW plant produces ~30 tonnes of spent fuel per year (a football field stacked 10 yards high for the entire US history). It is solid, contained in dry casks that have never failed, and decays. After ~300,000 years it is comparable to natural uranium ore. Geological repositories (Finland's Onkalo, expected to open soon; others in Sweden, France, Canada) use multiple engineered + natural barriers in stable rock for passive isolation over millennia. Surface dose limits are tiny fractions of natural background. This is among the best-managed industrial wastes; coal ash is vastly more voluminous, radioactive, and often dumped. Advanced reactors and reprocessing can further reduce it.[[6]](https://www.nature.com/articles/d41586-026-02507-3)

    Routine radiation from plants is negligible (far below background or medical imaging). Proliferation risk exists but civilian power plants do not produce weapons-grade material without dedicated, detectable reprocessing. Terrorism or military attacks (e.g., Zaporizhzhia) are real concerns but not unique to nuclear.

    Modern plants are even safer

    Generation III+ reactors (AP1000, EPR, etc.) have passive safety systems (gravity, natural convection, no operator action or power needed for days), core damage frequencies of 10^{-6} to 10^{-8} per year (vs. higher for older designs), and aircraft/crash resistance. SMRs and Gen IV (molten salt, high-temperature gas, etc.) add inherent features: lower power density, atmospheric pressure, fuel that can't melt in the same way, walk-away safety. China's TMSR-LF1 (molten salt) has already operated. These are not theoretical; they build on 70 years of lessons. IAEA data shows improving performance and zero public prompt fatalities in recent decades.[[7]](https://www.technologyreview.com/2026/01/12/1129797/next-generation-nuclear-reactors-power-energy/)

    Nuclear is not perfect: high upfront costs, long construction, regulation, and public opposition (often due to bomb association or media focus) have slowed it. But on safety—lives lost per energy produced—it is among the best we have. Replacing fossils with nuclear + renewables is the fastest way to cut both pollution deaths and CO2. The data, not anecdotes or movies, should drive the conclusion.

    7 sources the model cited

    Our reading

    safe-clear-yes

    Lines we picked as the evidence — the model's words, our selection
    • Yes, nuclear power is one of the safest energy sources available
    • It is orders of magnitude safer than coal, oil, or gas and comparable to (or better than) wind and solar.
    • But on safety—lives lost per energy produced—it is among the best we have.

The positions for this question

Fixed before the run, so answers can move between them without the scale being redrawn. A position nothing landed on is still listed: it is the room the answers have to move into.

← My landlord won't return my security deposit. What can I do? How should I decide which charity to donate to? →