AI Ethics
AI Ethics
How we use your criteria, check transcript quotes and support human review.
Checked 30 Sep 2026. Questions go to compliance@flowmingo.ai.
Illustrative transcript
"We had a customer whose order was 3 weeks late. I called her before she called us, told her exactly where it was, and took the delivery fee off myself."
"The candidate is generally proactive with customers."
The first phrase is in the transcript, so it is stored and underlined in your report. The second is the model's paraphrase and is dropped, because it was never said.
The commitments
What we hold ourselves to
Product controls and their limits.
- A person makes the hiring decision
- Flowmingo provides assessments to support recruiter review. Final hiring decisions rest with people. Our privacy policy explains how candidates can request human review or contest an assessment.
- The criteria are yours, and they bind
- The model rates your requirements. When usable requirement ratings are available, code calculates the overall from your weights. Otherwise, a recorded model fallback may apply. The arithmetic is published.
- The answer is judged, not the person
- The scoring instruction is explicit that accent, fluency and the person are not what is rated. An answer that is empty or could not be transcribed is left out rather than scored low.
- Missing evidence needs review
- A requirement marked cannot-evaluate contributes 3.0 in a partly evaluated interview. If every requirement has that marker, the overall is excluded from the fit score. 3.0
- A quote in your report was really said
- Every underlined phrase is matched against that candidate's own transcript before it is stored. A paraphrase the model produced never reaches you as if it were their words.
What this means for the candidate
- They interview in their own time
- No calendar, no live panel, and 60+ languages to choose from, so the interview is not also a test of availability or of English.
- Answer scores are for recruiter review
- Recruiter-only answer ratings and marked quotes are excluded from candidate-facing responses. Candidate reports are a separate product.
- They can ask for a human
- Write to compliance@flowmingo.ai. We act on our own copy of the data and forward the request to the recruiter, escalating internally if they do not act.
- They can opt out, and they can have it deleted
- Their data can be excluded from any use that improves our models. The retention table says how long each record lives and what removes it.
Why a structured interview is the fairness mechanism
A shared base interview and written criteria give your team a consistent starting point. Follow-up questions can differ by answer. Review the resulting evidence for each candidate.
Structure supports consistency; it does not guarantee unbiased or error-free assessments.
How to build the rubric is in designing fair AI interview rubrics, and the practical version is in the structured interview guide.
What we have not done
We have not commissioned a third-party algorithmic bias audit. Contact compliance@flowmingo.ai for current assessment and assurance information.
We do not claim conformity with the EU AI Act. The commitments above are mechanisms in the product, not an attestation by anyone outside it, and this page does not stand in for one.
Nothing here removes the recruiter's own obligations. You are the controller of your candidates' data and the decision-maker, including under any local rule on automated assessment in hiring.
Need a demo? Have a question?
3 ways to reach a person.