Products

Founded

Independent, cross-model verification of AI reasoning, purpose-built for the decisions with no ground truth to check them against.

Product Overview

Agreement alone isn’t proof. Founded is designed to test whether an answer holds up when it is deliberately challenged, so a result that merely looks unanimous isn’t mistaken for a trustworthy one.

Built for high-stakes judgment calls

Think contract interpretation, clinical decision support, credit and risk assessment, and compliance reviews. These are decisions where a confident but wrong AI answer is costly and there is no answer key to check against.

How it fits your workflow

  • Founded runs alongside the AI model you already use. It doesn’t replace it.
  • You decide which queries count as high-stakes. Only those are routed through Founded, and the answer is held until the check completes.
  • The model that produced the answer takes no part in checking it. Verification is done by independent models from other providers.

What you get back

  • A distinct, plainly labelled result for each evaluation, rather than a single confidence score. This includes results for when the evidence behind an answer doesn’t support it, or when the check couldn’t be completed.
  • An evidence trail of how each result was reached, to support your own governance and audit processes.

Status

Founded is in build, and we are seeking pilot clients for a future pilot program.

Patent pending AU2026907731

Reasoning divergence is when an AI’s stated reasoning, the evidence it was given and its final answer stop lining up.

What is reasoning divergence? →

← All products

We are actively seeking pilot clients.

If your organisation faces these problems today, we would like to hear from you.

Become a pilot client