How Clause Risk Scoring Would Actually Work

Not as a dedicated, repeatable tool — and we won't pretend otherwise. @vustbot (general multi-model AI chat) can already discuss a clause you paste and tell you whether it looks risky, free, in Telegram. What doesn't exist yet is a structured pass that classifies every clause in a document and tiers each one High/Medium/Low the same way every time. The waitlist button on this page is a real demand vote for that dedicated version.

Four steps: intake the full clause text (not a summary of it), classify each clause into a known category (termination, liability, IP, indemnity, non-compete, and so on), tier each one High/Medium/Low against common patterns for that clause type, then return a short scored summary instead of a wall of commentary. The mechanics matter because a risk checker that just reads and reacts gives inconsistent results — tiering against a fixed rubric is what makes it repeatable.

It means the clause deviates from what's typical for that contract type in a way worth a second look — asymmetric termination rights, an unusually broad IP assignment, an indemnity that runs one direction only. It is a heuristic flag, not a legal certainty. A High tier is a prompt to read that specific clause carefully or ask a question about it, not proof that it will cause harm.

No, and this matters. A certified legal risk assessment weighs your jurisdiction, your specific situation, and case law a lawyer knows and an AI checker doesn't reliably access. A tiered risk pass is pattern-matching against common contract norms — useful for triage, not a substitute for that judgment. For a real dispute, a large deal, or anything you can't walk back, a qualified lawyer's assessment is the one that counts.

Same underlying idea, different lens. /ai-contract-reviewer describes the overall output — a plain-English summary plus flags plus missing protections. This page zooms into the mechanics of just the risk-scoring step: how clauses get classified and tiered, and why tiering beats a flat list. Both point at the same waitlist; a shipped version would likely fold this scoring step into the general reviewer's output.

The intended design is general-purpose — services agreements, vendor contracts, leases, employment offers — because the underlying clause categories (termination, liability, IP, indemnity) recur across contract types even though the specific wording differs. NDAs get their own dedicated angle on our free NDA analyzer page, since NDA risk lives in a narrower, more specific set of variables (direction, duration, non-solicit scope) than a general risk pass covers.