Why this exists
Most AI-risk coverage mixes three very different kinds of claim in one sentence: something that actually happened today, a number from a dataset that was accurate six months ago, and someone's guess about 2029. Read fast enough, they all sound equally certain. That's the problem this dashboard is built around — not adding another risk score to the pile, but refusing to let those three blur together.
So every figure on the dashboard carries one of three tags, and the tag isn't decorative. means it's polled from the source RSS feeds every few minutes. means it's real and cited but dated — the AI data-center map comes from Epoch AI's published dataset, not a live crawl. The map's Network view follows the same rule: every line between a lab, a data center and its owner is a relationship Epoch actually lists, the ones Epoch rates as only likely or speculative are drawn dashed, and sites with no listed owner or user are left out rather than guessed at.
means it's a hand-built toy model — useful for seeing how a threat score might move if you drag the Time Machine forward, useless as a prediction, and never presented as one.
That last category is the one worth being honest about. The threat gauge, the capability-vs-safety curve, five of the seven Risk Radar axes — none of it is a benchmark, an evaluation of any named lab or model, or a forecast. It's a model I built to make a shape visible: what it looks like, roughly, if capability keeps compounding faster than the safety work meant to keep up with it. The shape is the point, not the digits after the decimal.
The two exceptions are Energy & Resource Strain and Compute Monopolization, which are real arithmetic on the data-center snapshot (aggregate IT power against Epoch's own 2030 estimate, and a Herfindahl–Hirschman index over who owns it) rather than anything hand-set. Everywhere else, if it's marked simulated, treat it exactly like that — a toy, not a source.