What this tool is

An aggregation-and-calculation tool. It takes real, cited published figures and makes them interactive. You tell it company size and use case, it returns a plain-language estimate of both compliance cost and exposure, with every number linked to a public source.

What this tool is not

Key design decisions

SME / large-undertaking distinction

Article 99(6) is the genuinely non-obvious detail: for SMEs including startups, the fine is the lower of the fixed cap and the percentage of turnover. For large undertakings, the fine is thehigher of the two. This is the opposite rule and it changes the answer materially, so the calculator shows both binding candidates side-by-side.

Ranges with citations, not point estimates

The cost side of the ledger genuinely is a range — different use-case shapes cost different amounts, and the underlying studies (CEPS, Holistic AI, SQ Magazine, Ovidiu Sciu) report ranges. Picking a midpoint and presenting it as if it were a quote would mislead founders, so we don't. Every number on the page is the underlying range with a citation chip.

Why Colorado is qualitatively, not quantitatively, represented

Colorado's SB 26-189 enforces via the state's Consumer Protection Act as a deceptive-trade-practice claim, prosecuted by the AG. There is no published AI-specific fine schedule. Forcing it into the same euro-figure format as the EU numbers would be inventing a number that doesn't exist, which violates the tool's sourcing rule. The Colorado card shows the enforcement model, the cure window, and the primary statute instead.

Why the calculator runs client-side

The calculation logic is a handful of lookups against the penalty-tier and cost-range tables. It runs entirely in the browser — no backend, no serverless function, no API route. This means no data leaves the browser, no Vercel function-duration limits to think about, and the full source data can be a small typed TS config file where updating a figure is a one-line edit.