How it works
After Phase 2 dimension scoring, the Integrity Screen evaluates whether material governance failures should cap the maximum achievable grade.Five concern categories
| Category | What it evaluates |
|---|---|
| Safety incidents | History of AI safety failures, unmitigated risks, or inadequate incident response |
| Ethical violations | Documented bias incidents, privacy breaches, or violations of stated AI principles |
| Regulatory actions | Enforcement actions, fines, consent decrees, or ongoing investigations related to AI |
| Governance failures | Board-level failures: lack of oversight, ignored warnings, or governance structure collapse |
| Stakeholder harm | Documented harm to employees, customers, communities, or other stakeholders from AI systems |
Ceiling logic
Each concern category, when triggered at a critical level, imposes a maximum grade ceiling:- Critical failure in one category: Maximum grade capped at A (cannot achieve AA or AAA)
- Critical failures in two categories: Maximum grade capped at BBB
- Critical failures in three or more categories: Maximum grade capped at B
- Pattern of unaddressed critical failures: Grade floor of CC or below
The ceiling applies to the final grade, not individual dimension scores. Dimension scores remain as calculated in Phase 2, providing a clear picture of where governance is strong even when the overall grade is capped.
Remediation path
Ceiling caps are not permanent. Companies can address Integrity Screen findings through:- Documented remediation of the specific failure
- Structural changes to prevent recurrence (not just policy updates)
- Independent verification that remediation is effective
- Sustained track record (typically 12-18 months) without recurrence