INSIGHTS · TALENT RISK · NO. 001

We Interviewed 20 Candidates. 17 Weren’t Who They Said They Were.

One search exposed how quickly AI‑enabled candidate fraud is becoming an operating and security risk.
Jason Crane  ·  AUGUST 2026  ·  7 MIN READ

The search itself was unremarkable. A senior engineering role. Remote‑friendly. Strong but standard compensation. The kind of mandate we run every month. Twenty candidates reached the interview stage.

Seventeen of them were not who they said they were.

17 / 20
candidates failed identity or experience verification. One search. One ordinary role.
What failing looked like

Failing does not mean a polished résumé or a generous title. It means the person, the experience, or both could not survive verification. Employment at real companies in roles that never existed. References who answered on the first ring, every time, with the same warm fluency. Technical depth that collapsed the moment a question left the script, because the answers were arriving from somewhere else.

And one candidate whose face did not move the way faces move.

The one we keep thinking about

His credentials were clean. His references were warm. On camera he was composed, articulate, credible. Then he turned his head, and his jawline lagged behind the turn. We asked him to lift a hand to his face. He hesitated in a way no honest person hesitates at that request, and the image seemed to negotiate with itself before settling.

He was altering his appearance in real time, during a live interview, competently enough that two experienced interviewers almost missed it.

We ended the call politely. We still do not know who we were talking to. That sentence should bother you as much as it bothers us.

This is not a recruiting problem

A bad hire costs you two quarters. A synthetic hire costs you whatever they came for. The person you hire receives credentials on day one: repositories, production systems, customer data, and the internal wiki that maps where everything valuable lives. If the identity is fabricated, you did not make a hiring mistake. You granted system access to someone you have never met.

Most verification still happens after the offer, through background‑check vendors built for a world where fraud meant an inflated title and a stretched date range. AI has collapsed the cost of manufacturing all of it at once: the résumé, the references, the portfolio, the face. The economics moved. Most processes have not.

If your process assumes the person on the screen is real, that assumption is now the weakest control you have.
What we changed

Live working sessions where candidates solve real problems with cameras on and hands visible. Document and liveness verification before final rounds, not after offers. References we source ourselves rather than dial from a supplied list. In‑person finals wherever the role touches production systems or customer data.

None of it is exotic. It is simply treating the interview as what it has become: an access decision.

Seventeen out of twenty was one search, one quarter, one ordinary role. The number that matters is yours.

CNA Search Insights  ·  Talent Risk, No. 001  ·  Client and candidates anonymized. Details altered only where required to protect identity.