Finding got cheap. Proving got hard. Scanners, AI review tools and outside reports produce more candidate findings than a team can judge. We take the findings you already have, check each one against your code, and tell you which are confirmed, which are refuted and which no check has settled.
When you need it
- A tool has produced more findings than your team can judge before a release.
- A penetration test or a bounty report says “critical” and your engineers disagree.
- A decision on a finding will have to be justified later, to a customer, an insurer or an acquirer.
What you send us, once the confidentiality agreement is signed
- Up to five findings, in the form you received them, and the name of the tool, firm or person that raised each one.
- The code they refer to, at a commit you name.
- The decision the result has to support, and your deadline.
What you get
A signed report under a persistent identifier (CPR-YYYY-NNN). For each finding it gives one decision:
- Confirmed, with the evidence: file and line, steps to reproduce it, and the evidence rung it reached.
- Refuted, with the reason.
- Plausible only, where no check settled it, with what was checked and what would settle it.
Illustrative specimen, not from a client engagement; identifiers are placeholders.
What a decision does not mean
- “Refuted” is about the finding as you submitted it, at that commit. It does not mean the software is free of that defect in another form, or of any other defect.
- An adjudication is not a review. We examine the findings you send, not the rest of your code. If we notice something else on the way, it goes into the record you receive, marked as outside the submitted findings.
- We do not write or merge the fix. That separation is a condition of the review’s independence.
Why ask us
We study what agreement between AI models is worth, and we publish it. A model checking its own work is not an independent review, and models from different companies are not automatically independent of each other. So each AI reviewer’s output is written down before anyone compares them, findings do not close by vote, a named person signs the decision, and we withdraw it in writing if it does not survive.
What you should know first
CounterProof is part of a group that builds payment and digital-asset custody infrastructure. If you build in those markets, our affiliate may be adjacent to you or competing with you. We tell you in writing what the group builds and where it operates before any engagement and before any money moves. We issue no assessment on code authored by CounterProof or any company in our group.
How long, and what it costs
A fixed fee, quoted in writing after a scoping call. It does not depend on how the findings come out. For up to five findings we aim to deliver within 10 business days of receiving access.
What we take on
We take on findings in any technical area. Where an area is new to us, we can first study it by our method and build a reference library from public sources. That study is free of charge and commits you to nothing. If we cannot judge a finding properly, we say so and decline it before anything is signed.
How to start
Email us how many findings you have, what kind of source raised them, and your deadline. Please do not send the findings or any code yet. Those follow once the confidentiality agreement is signed, and until the scope is agreed we read them ourselves, without any AI provider. Elenora Soons, our account manager, will arrange a call with the review team.