Which AI could end humanity first?
Nobody can measure that — not this site, not anyone. There is no dataset of futures, so any ranking of real products by predicted lethality is an opinion wearing a number.
Here is the neighbouring question that is measurable, and that this page actually answers: which AI is the internet most afraid will? That is a fact about what people say, and it can be counted, weighted and audited.
The internet's answer this week
Anthropic
Anthropic · 374 documents in 168h
23.4
Doom Index / 100
9h 54m to midnight
on the Doom Clock
It leads chiefly on existential-risk talk — not because anything about it is more dangerous, but because that is what the corpus contains. The breakdown below shows every entity's three terms separately.
Ranked by how much the internet fears them
Doom Index over 168h. Measures existential-risk discourse, not risk. · Higher means more doom talk
Why each one is where it is
The Doom Index is three measurements added together. These are their contributions in index points. · Segments sum to that entity's score
- Existential share (×0.55)
- Fear intensity (×0.25)
- Who is talking (×0.2)
What would move this ranking
Read this next
- The Doom Clock — the same measurement as one number over time
- The composite index — fear and trust across all AI discourse, not just existential
- Methodology — every formula, and what each one cannot tell you