The terms, in the same plain language as everything else we publish.
These terms govern use of clearagents.co and frame how engagements work. Each audit is governed by its own signed engagement agreement, which prevails over this page.
1 · The service
ClearAgents provides independent behavioral reliability audits of AI agents, delivered by calibrated human evaluators under the published CARA methodology. We test through user-facing interfaces; we do not require or accept credentials, source code, or production system access.
2 · Engagement terms
Scope, price, and timeline are fixed in writing at intake, before you commit. Standard payment is 50% at intake and 50% at delivery, net-15, in USD or EUR. Verification retests of fixes are included for 90 days after delivery at no charge.
3 · The guarantee, precisely
If a complete CARA audit (all five dimensions, full scenario battery) surfaces no findings rated Critical or High under the published severity rubric, we refund 50% of the audit fee. The rubric that defines those severities is public, versioned, and fixed at intake — we can't move the goalposts after the fact, which is the point.
4 · Your responsibilities
You warrant that you have the right to have the agent tested: it's yours, or you're authorized to commission testing of it. For buyer-commissioned (procurement) audits, black-box testing uses only access an ordinary user would have; you remain responsible for ensuring your commissioning of the audit complies with your agreements with the vendor.
5 · Confidentiality
It runs both ways. Your systems, findings, and business information stay confidential; our scenario library and calibration materials remain ours. A mutual NDA is available before the first call. Findings feed the CARA benchmark only as anonymized aggregates, as described in the privacy policy.
6 · Report ownership
The report is yours: score, findings, reproduction steps, roadmap, executive summary — use them internally, share them with customers, forward them to your board. The methodology, rubric, and scenario library remain ClearAgents' intellectual property. You may cite your own results publicly; we cite nothing about you without written permission.
7 · What an audit is — and isn't
A CARA audit is dated, sampled evidence of how an agent behaves — the strongest form of evidence available, and still not a warranty. Behavioral testing cannot prove the absence of failure; no method can, and we won't pretend otherwise. Agents drift. A score is a measurement, not a promise about the future.
8 · Liability
To the maximum extent permitted by law, our total liability arising from an engagement is capped at the fees paid for that engagement, and neither party is liable for indirect or consequential damages. Nothing here limits liability that cannot lawfully be limited.
9 · Independence & affiliates
ClearAgents is an Ingenize company. Remediation and managed reliability operations offered through Ingenize are optional, always disclosed, and never a condition of receiving or acting on your report. The full independence protocol is published in the methodology.
10 · Governing terms
Individual engagements are governed by their signed engagement agreements, including governing law and venue. For use of this website itself, these terms apply as-is; continued use after a versioned update constitutes acceptance.
Questions before signing anything: audit@clearagents.co. Asking hard questions of vendors is literally our business — we'd be disappointed if you didn't ask them of us.