B · PlausibleTechnology

Advanced AI risk

Some researchers argue that increasingly capable AI systems could cause catastrophic harm, through misuse or through systems that pursue goals humans did not intend. Others consider this speculative. There is no live data to track, so this page only shows what experts say, and how much they disagree.

The tier rates the quality of the evidence, not how likely the event is.

How it could unfold

1AI capability grows quickly◐ moderate evidence
2Misuse, accidents or misaligned goals cause serious harm○ speculative
3Harm at a scale that is hard to reverse○ speculative

Historical anchors

None for the extreme outcomes. Smaller harms (misinformation, fraud, automation) are already visible.

What experts and models say

Source	Statement	Kind
Toby Ord ↗	1 in 10 chance, per century, of an existential catastrophe from unaligned AI (his own estimate, the highest in his table)	EXPERT ESTIMATE
Grace et al. (2024) ↗	In a survey of 2,778 AI researchers, between 38% and 51% (depending on the question) gave at least a 10% chance to outcomes as bad as human extinction; answers varied widely	EXPERT ESTIMATE

Expert estimates are personal judgements, not measurements, and experts disagree. This site never adds its own probabilities.

Caveat

These are opinions, not measurements. Wide disagreement among experts is the main finding.

What you can do

Nothing practical at household level. Following the policy debate is the main way to engage.

Open the preparedness checklist →

Related scenarios

Engineered pandemic

References

Ord, T. (2020). The Precipice: Existential Risk and the Future of Humanity, Table 6.1. Hachette. link ↗
Grace, K. et al. (2024). Thousands of AI authors on the future of AI. arXiv:2401.02843. link ↗