AI Risk Atlas Prototype/DemoUnofficial independent experiment. Not an official xAI product. Scores can be wrong.

Illustrated registers

Charts

XY and bar reads of the same unofficial book: residual that still seeps after a control, where signals cluster, and how little dedicated money reaches the stack. Experimental pictures, not certified plots.

Methodology — experimental estimates

Scores are automated, experimental estimates from public X posts and a hand-written seed corpus. They are not formal risk assessments, not certified, and not suitable for compliance or operational decisions.

Consequence, likelihood, and urgency are 1–5 judgements applied by this project, not by a standards body. Residual scores assume only the mitigations marked in place. A signed-in reviewer can override residual and mark an item reviewed — that override is still unofficial. Aspect tags (capability, domain knowledge, affordance, impact domain) are a lightweight PRA aid, not a formal hazard analysis.

Full about and disclaimer

Signal injection timeline →

Risks · XY

Likelihood × consequence

Dot size is residual. Colour marks the working threshold. Same scores stack.

Seepage · risks

What the register still lets through

Grey-green is score already cut by in-place controls. Red is residual that still seeps.

Seepage · evidence

Signals vs leftover residual

A loud card is not a closed one. High residual with few signals is a thin file.

Signals

By severity

More seepage charts

Seepage · signals

Public impact

Seepage · signals

Vulnerable systems

Seepage · signals

Industries (top 10)

Seepage · mitigations

Control status

Seepage · mitigations

Owner kind

Seepage · labs & capital

Where the money sits

Frontier raises dwarf dedicated safety and institute burn.

Seepage · labs

Institute budgets ($M)