Free tools Windows power users keep installed
One-click scans. No signup required.
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
AI could cause severe harm, but the evidence does not justify saying that an apocalypse is inevitable—or that human extinction is impossible. The useful way to assess the danger is to separate harms already being observed from future loss-of-control scenarios, and to judge each by likelihood, severity, evidence, time horizon, reversibility and governability.
How to read this AI risk scorecard
“AI apocalypse” can mean very different things: misinformation spreading at scale, people losing work, malicious use of AI, systems escaping meaningful human control, or human extinction. Treating those outcomes as one risk obscures how much evidence exists for each and what can be done about it.
This scorecard uses six questions. The ratings are qualitative assessments of the evidence described below, not numerical forecasts or official ratings.
- Likelihood: How plausible is the pathway given the available evidence?
- Severity: How large could the harm be?
- Evidence quality: Is the concern grounded in observed harms, empirical assessment or a hypothetical scenario?
- Time horizon: Is it happening now, a near-term concern or a longer-term possibility?
- Reversibility: Could the harm be contained or undone?
- Governability: Are practical controls, monitoring and accountability available?
A high severity rating does not mean a scenario is likely. Likewise, uncertainty about a future catastrophe is not evidence that it will happen—or proof that it cannot.
#1 Best Overall
AI risk scorecard
| Risk pathway | Likelihood and evidence | Severity and horizon | Reversibility and governability |
|---|---|---|---|
| Misinformation and disinformation | Current and observable; the World Economic Forum identifies it among major AI-related global risks. | Potentially broad social harm; current risk. | Some individual false claims can be corrected, but spread and effects may be difficult to reverse. The concern is governable in part through monitoring and accountability, though no single control resolves it. |
| Job loss and displacement | A current distributional concern; the World Economic Forum highlights displacement risks. | Can affect livelihoods and how economic gains are distributed; current risk. | Impacts may be mitigated, but a displaced worker’s losses are not automatically undone. Governance depends on how organizations and policymakers manage deployment and its effects. |
| Malicious enablement | The International AI Safety Report and World Economic Forum discuss pathways including AI lowering barriers to malware and fraud, and potentially biological misuse. | Potentially serious, with the scale depending on the misuse and its consequences; a present concern and an evolving pathway. | Prevention and containment are possible areas for controls, including restricting access to dangerous capabilities. Whether particular measures work depends on the system and threat. |
| Loss of control | A future scenario, not an established outcome. The International AI Safety Report discusses uncertainty around systems outmaneuvering human operators. | In severe cases, the report says outcomes could include human marginalisation or extinction; the horizon is uncertain and future-facing. | Potentially difficult to reverse if people lose effective control. Governability hinges on evaluation, access controls, monitoring and the ability to intervene. |
| Human extinction | Unresolved tail risk. RAND’s 2025 analysis says creating an extinction threat would be immensely challenging, but cannot be ruled out. | Maximum-severity outcome; a hypothetical future pathway, not a documented consequence of current AI. | Irreversible by definition. Preventive governance matters because response after such an outcome would not be possible. |
| Power concentration and rights | The United Nations AI governance process treats coordination, accountability and equitable participation as central concerns. | Can affect rights and who holds influence over AI; a governance concern across deployment and development. | Potentially addressable through accountability, liability and participation in decision-making, though implementation requires coordination across jurisdictions. |
Is AI really going to cause an apocalypse?
No one can responsibly answer that with certainty. The evidence supports concern about present harms and serious attention to future catastrophic scenarios; it does not establish that extinction or another “apocalypse” is inevitable. The distinction matters: observed misinformation or displacement should not be treated as proof of an extinction pathway, while the speculative nature of extinction risk does not make it safe to ignore.
RAND’s 2025 analysis puts the uncertainty plainly: creating an AI-driven extinction threat would be immensely challenging, but cannot be ruled out. That is not a probability estimate. It is a statement that the pathway is difficult and uncertain, yet not conclusively excluded.
Rank #2
What do AI researchers think about the odds?
In the 2024 AI Impacts survey of 2,778 AI researchers, 68.3% judged good outcomes from superhuman AI more likely than bad. At the same time, many respondents assigned at least a 5% chance to extremely bad outcomes.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteThose findings describe researchers’ judgments, not a measured frequency or a consensus forecast that extinction has a particular probability. The two results can coexist: someone may consider good outcomes more likely overall and still regard a low-probability catastrophic outcome as important enough to address.
What is already happening, and what remains hypothetical?
Documented or present-day concerns
Misinformation and disinformation, job displacement, and malicious use belong on the present-risk side of the scorecard. The cited sources identify them as meaningful concerns; the evidence in this scorecard does not quantify their frequency, the amount of harm attributable specifically to AI, or how those harms compare across regions.
Future catastrophic scenarios
Loss of control and human extinction are different kinds of claims. The International AI Safety Report discusses severe loss-of-control outcomes as uncertain scenarios. RAND’s 2025 assessment says an extinction threat would be immensely challenging to create but cannot be ruled out. Neither claim establishes that current systems are on a demonstrated path to extinction.
The Center for AI Safety statement, signed by hundreds of researchers and technology leaders in 2023 and reproduced in the International AI Safety Report, says: “Mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war.” It is a call to prioritize risk mitigation, not evidence that the risk is certain or imminent.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What can governments and AI developers do?
Governance is part of the risk assessment, not an afterthought. A hazard is more governable when organizations can evaluate systems independently, limit access to dangerous capabilities, report incidents, and be held accountable when controls fail. The OECD’s 2024 policy assessment points to clearer liability rules, AI “red lines,” investment in AI safety and adequate risk-management procedures. The United Nations’ 2024 “Governing AI for Humanity” report offers an internationally consulted governance blueprint.
Best Value
Benchmarks for a credible response
- Independent evaluation: Test consequential capabilities and risks outside the team or organization building the system.
- Incident reporting: Make serious failures and misuse visible enough for organizations and authorities to respond.
- Model access controls: Limit access where a system’s capabilities could enable serious misuse.
- Liability: Clarify who is accountable when development or deployment causes harm.
- Red-line prohibitions: Define uses or capabilities that should not be deployed.
- Cross-border coordination: Align reporting and safety expectations where risks and systems cross national borders.
These are practical indicators to look for, not a guarantee of safety. A policy counts only if it is implemented, monitored and enforceable; international agreement also has to translate into action by governments and AI developers.
How much should you worry?
Worry should track both evidence and stakes. Present harms merit attention because they are observable and can affect people now. Catastrophic loss-of-control and extinction scenarios warrant serious prevention because their potential severity is extreme, even though the pathways remain uncertain. The most defensible position is neither certainty of doom nor certainty of safety: distinguish the risks, demand evidence proportional to the claim, and judge institutions by whether they can detect, limit and take responsibility for harm.
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.
