Snapshot · 21 Sept

Shared ranking window · 3 days

Every AI system we track, and what people are saying about it

This is not a ranking of which AI is best. It is a measurement of where public discourse currently sits for each one, on a scale from trust at 0 to fear at 100. A high number means people are talking about it fearfully — not that it is dangerous, and not that it is bad.

Overall index

56.6

fear

Models tracked

3

ranked separately from labs

Labs tracked

2

not a total of their products

No single theme

3

of 5 — the four fears are evenly split

All of them, side by side

Models and labs are listed together here for scanning only. They are ranked separately below, because a product and the company that ships it are not competitors: what is said about the product is also part of what is said about the company.

#EntityIndexReadingDiscussed asFear / trustSample
1OpenAIlab54.5Neutralno single theme26 / 1207
2Geminimodel53.6NeutralExistential risk45 / 266
3Anthropiclab53.1NeutralExistential risk24 / 7235
4Claudemodel50.4Neutralno single theme25 / 1844
5ChatGPTmodel49.3Neutralno single theme22 / 3159

Models & products

3 ranked
1Model

Gemini

53.6Neutral

Discussed mainly asExistential risk

fear mass 44.7%trust mass 1.5%polarisation 7n = 66
2Model

Claude

50.4Neutral

Discussed mainly asno single theme(lead 1.5 pts)

fear mass 24.8%trust mass 17.6%polarisation 83n = 44
3Model

ChatGPT

49.3Neutral

Discussed mainly asno single theme(lead 2.8 pts)

fear mass 22.2%trust mass 31.0%polarisation 84n = 59

Companies & labs

2 ranked
1Lab

OpenAI

54.5Neutral

Discussed mainly asno single theme(lead 3.7 pts)

fear mass 26.1%trust mass 1.2%polarisation 9n = 207
2Lab

Anthropic

53.1Neutral

Discussed mainly asExistential risk

fear mass 23.7%trust mass 7.2%polarisation 47n = 235

The four things people are afraid of

Every entity above is scored against these four. When no single one leads by more than the sampling error, the entity is marked as having no single theme rather than being given the largest of four near-equal shares.

Existential risk51.5

n = 74

Jobs & economy50.0

n = 29

Deepfakes & deception50.8

n = 27

Privacy & data52.5

n = 25

Every figure here is a measurement of discourse, not an assessment of a product. Read how it is calculated, or open the full dashboard for the charts, the confidence intervals and the source mix.