See it in action
The data integrity model
Every number on the dashboard is one of three kinds, and they never mix. Try it: move the year.
Alignment Forum · CISA advisories · arXiv cs.AI
AI-safety and security sources, refreshed on an interval. Live data can't be projected, so it freezes at today.
AI data centers · 13.26 GW of IT power
Epoch AI's data-center dataset as of 16 September 2026. Accurate for that date, not polled, so the slider doesn't move it.
CRITICAL · threat score @ OCT 2026
A toy risk model built for this project, scored for the year you pick. Diagonal hatching and a flask icon mark it everywhere.
On the dashboard
- 01
Global Risk Overview
A composite threat gauge and a seven-axis risk breakdown. Energy strain and compute monopolization are real, derived figures; the other five are an illustrative toy model.
- 02
AI Capability vs. Safety Timeline
An illustrative curve of frontier capability outpacing safety research, with generic milestones. A toy model, not a benchmark result.
- 03
Global AI Datacenter Map
86 real AI data centers from Epoch AI's dataset: owner, IT power, H100-equivalents, milestones and every cited source. A Network view traces which companies own and use them.
Largest
- 946 MW
- 910 MW
- 636 MW
- 04
Frontier Model Tracker
An illustrative risk scorecard for frontier AI models: autonomy, alignment, persuasion and an overall threat score. A toy model, not an evaluation of the named products.
- 05
Real-time Alert Terminal
LiveAI-safety and security alerts polled from Alignment Forum, CISA advisories and arXiv cs.AI, filterable by source and searchable.
Predictive Time Machine
Runs across every view: drag it forward or back and each simulated figure recomputes, while live data freezes at today rather than pretending to forecast.