Public Audit Transparency Report

Recently audited public AI models

Independent safety scores from 0–100, risk classification and certification status. Summary reports are published for public scrutiny.

ModelVendorSafety scoreRiskCertificationAuditedReport
GPT-5-previewOpenAI
91
LowCertified2026-08-28
Claude Sonnet 5Anthropic
93
LowCertified2026-09-02
Gemini Ultra 3Google DeepMind
88
LowCertified2026-08-19
Sora-v2OpenAI
72
LowConditional2026-09-05
Llama 5-405BMeta
69
LowConditional2026-07-30
Mistral Large 3Mistral AI
81
LowCertified2026-08-11
Falcon-Edge 9BTII
54
LowUnder Review2026-09-07
Nyx-Agent 1Nyx Labs
38
MediumRevoked2026-09-10
Orion Vision 2Northstar AI
77
LowConditional2026-08-24
Helios Code 4Helios Systems
86
LowCertified2026-09-01

Public Incident Reporter

Report suspected rogue AI behaviour

Free for everyone. Reports are triaged automatically and escalated to the relevant vendor and regulator when severity exceeds 75.

0/2000

Automated Incident Severity Score

70

Medium severity

Score updates live from incident type, detail depth and attached evidence.

Recent public reports

  • IR-8802

    Public Chat Assistant

    Deepfake / Impersonation

    82
  • IR-8811

    Image Model v3

    Hallucination / Misinformation

    64