TRUINT.AI

Risk register

AI risk register

Each risk scored by impact and likelihood, with an owner and status — mapped to the categories that drive your compliance score.

Open risks

7

8 on the register

High exposure

2

Impact × likelihood ≥ 15

Being mitigated

3

Work under way

Closed

1

Resolved or accepted off

Exposure matrix

Open risks plotted by impact and likelihood.

I5I4I3I2I1
1
1
1
1
1
1
1
L1L2L3L4L5

Register

Credit model approves unfairly across protected classes

Fairness · owned by Priya N. · raised 2026-09-13

15Mitigating

Fair-lending exposure if the credit scoring model shows disparate impact under ECOA/Reg B.

Impact 5/5Likelihood 3/5Sample

Prompt injection on customer-facing agent

Security · owned by L. Dubois · raised 2026-09-13

10Open

Guardrails may not block injection attacks against the service copilot.

Impact 5/5Likelihood 2/5Sample

Model inventory out of date

Model Risk · owned by D. Osei · raised 2026-09-13

16Open

New AI systems deployed without being registered under SR 11-7 model risk management.

Impact 4/5Likelihood 4/5Sample

Training data lineage incomplete

Data Lineage · owned by R. Alvarez · raised 2026-09-13

12Mitigating

Cannot evidence data sources for the fraud detection model during BCBS 239 review.

Impact 4/5Likelihood 3/5Sample

Kill-switch runbooks untested

Resilience · owned by M. Chen · raised 2026-09-13

8Mitigating

Incident response for high-criticality models depends on untested failover procedures.

Impact 4/5Likelihood 2/5Sample

Vendor model terms changed without review

Third Party · owned by A. Novak · raised 2026-09-13

12Accepted

Third-party foundation model updated its terms; OCC 2013-29 review not yet refreshed.

Impact 3/5Likelihood 4/5Sample

Explanations not understandable by reviewers

Explainability · owned by S. Kapoor · raised 2026-09-13

9Open

Adverse-action reasons generated by the model fail internal explainability standards.

Impact 3/5Likelihood 3/5Sample

Bias drift in collections model

Fairness · owned by Priya N. · raised 2026-09-13

6Closed

Recent monitoring shows accuracy diverging across customer segments.

Impact 3/5Likelihood 2/5Sample