← Open interactive AI Risk Trajectory

DATASET · 2026-09-17

The data behind the visualisation.

The interactive product is a view onto a versioned evidence model. The graphical interface is not the source of truth. The baseline now separates top-level systems, specific pathways and claim-level evidence.

7risk systems
30monitored pathways
32claim-level evidence records
13control records
10core source records
6evidence gaps

Assessment records

Autonomy & human control: R2 / X2 / C6 / A2

Cyber & digital security: R3 / X3 / C4 / A2

Biological, chemical & physical harm: R2 / X2 / C5 / A2

Information & human agency: R3 / X3 / C4 / A1

Critical systems & systemic dependence: R1 / X3 / C4 / A1

Economic & financial systems: R1 / X3 / C4 / A2

Geopolitical & strategic systems: R3 / X3 / C5 / A1

Evidence architecture

Sources are documents or datasets. Evidence claims record the specific proposition a source is being used to support, limit or contextualise. Pathways group claims into causal mechanisms. Systems summarise the furthest validated pathway state without implying every pathway has reached that state.

Versioning

Methodology version, evidence snapshot and software release are tracked separately so interface changes cannot silently alter scientific state. The baseline contains 6 material timeline events. Read the methodology or the current AI Risk Pulse.