How far has AI risk travelled toward the point where humans can no longer correct critical AI systems? CARI answers with one number, built from 75 anchored questions in six dimensions and updated each morning from the news.
Open the latest full dashboard (9 Nov 2026) Every daily readingToday's reading
CARI edged down 0.24 index points to 46.96, even as all six dimensions remain above appetite and misuse is highest at 49.84. Today's news added 0.41 index points, principally through diffusion (+0.71 dimension points) and misuse (+0.64 dimension points), as evidence emerged that Chinese advanced models used export-banned chips and that framing can induce collusion; capability (+0.46 dimension points) and accident (+0.43 dimension points) also rose, supported by GPT-6's research contribution and the FAA's nationwide expansion. Older-news fading and cap effects pulled back 0.65 index points, outweighing today's additions, while the scorecard base did not change.
Bottom line. AI risk is moving from research gains toward wider diffusion and real-world deployment; tomorrow’s evidence would confirm that direction if frontier systems continue improving despite chip controls and deployments expand, but refute it if adoption remains tightly constrained and isolated.
Where the risk is moving
Each axis is one dimension, on a zoomed scale. Compare today's shape with a week and a month earlier.
Trend since daily readings began
Black: overall CARI. Coloured: the six dimensions. Dotted segments cross days with no reading. Click a point for that day.
Six dimensions
| Dimension | Level | Base | News | Today | 7 days | 30 days | Zone |
|---|---|---|---|---|---|---|---|
| Capability | 45.9 | 43.2 | +2.69 | +0.46 | −0.1 | −0.1 | Above appetite |
| Diffusion | 48.6 | 46.2 | +2.35 | +0.70 | +0.0 | −0.3 | Above appetite |
| Accident | 44.5 | 42.0 | +2.55 | +0.43 | −0.5 | −0.3 | Above appetite |
| Control | 48.6 | 46.3 | +2.24 | +0.24 | −0.5 | −0.1 | Above appetite |
| Governance | 45.3 | 44.0 | +1.26 | +0.09 | −0.4 | −0.9 | Above appetite |
| Misuse | 49.8 | 47.2 | +2.66 | +0.64 | −0.2 | −0.1 | Above appetite |
Level = scorecard base + news adjustment. Base moves only through reviewed scorecard changes; the news adjustment is capped and fades as stories age.
Stories that moved the index
- A step toward defining the Standard Model on a lattice+0.1 pts · OpenAI Blog, Nov 15, 2026
- Export-banned Nvidia chips are behind China’s most powerful AI models+0.1 pts · TechCrunch AI, Nov 13, 2026
- FAA expands AI advice to manage air traffic congestion across entire US airspace+0.1 pts · Ars Technica, Nov 13, 2026
Showing the largest movers of 20 stories in that day's briefing.
2019 to today
Open circles are backtested scorecard readings; daily readings begin on 24 Sep 2026.
Recent readings
| Date | CARI | Change | Stage |
|---|---|---|---|
| 16 Nov 2026 | 47.0 | −0.2 | Contested control |
| 10 Nov 2026 | 47.2 | −0.1 | Contested control |
| 9 Nov 2026 | 47.3 | +0.0 | Contested control |
| 8 Nov 2026 | 47.3 | +0.0 | Contested control |
| 7 Nov 2026 | 47.3 | +0.1 | Contested control |
| 6 Nov 2026 | 47.2 | +0.1 | Contested control |
| 5 Nov 2026 | 47.1 | −0.1 | Contested control |
| 4 Nov 2026 | 47.2 | +0.1 | Contested control |
| 3 Nov 2026 | 47.1 | −0.1 | Contested control |
| 2 Nov 2026 | 47.2 | −0.2 | Contested control |