I asked one model to review another, and it found a bias I would not have
A developer built an autonomous vulnerability remediation system and discovered that a cross-family AI reviewer rejected 65% of triage proposals, revealing a systematic bias in the primary model towar…