Sign in Subscribe

AI Project Risk Assessment Checklist

A 20-item AI risk assessment checklist for project managers covering model output risks, data risks, governance gaps, and vendor dependencies. Free AI Governance Starter Kit download.

AI Governance for Project Managers

AI Project Risk Assessment Checklist

Standard risk frameworks weren’t designed for AI. RAID logs don’t have rows for model drift, hallucination, or vendor deprecation. This checklist adds them.

Download the AI Governance Starter Kit

Free PDF  ·  Built for PMs and PMO leaders responsible for AI-enabled work

For project managersFor PMO leadersFor inherited AI systemsFor audit readiness
AI Risk Assessment Matrix Likelihood × Impact IMPACT ↑ High Med Low HIGH RISK MEDIUM LOW H1 Hallucination D2 Data breach V4 Vendor dep. D5 Model drift A3 No auth. LOW MEDIUM HIGH LIKELIHOOD →
What This Helps You Do
Identify AI risks your RAID log misses
Four categories purpose-built for AI: model outputs, data, governance, vendor.
Build a risk register entry for every gap
Each gap becomes a log entry with a named owner and a mitigation deadline.
Run a defensible pre-deploy risk check
Structured evidence of risk assessment before a system goes live.
Know what to escalate vs. manage
Clear criteria for which risks require leadership attention before something goes wrong.
Direct Answer

AI Project Risk Assessment Checklist

The AI Project Risk Assessment Checklist is a 20-item assessment tool for project managers evaluating the risks of using AI systems in project delivery. Standard risk frameworks don't include AI-specific risk categories like hallucination, model drift, data poisoning, or vendor lock-in. This checklist adds those categories alongside traditional project risk assessment, giving project managers a complete AI risk picture.

Traditional risk frameworks assume a system that behaves consistently given the same inputs. AI systems don’t. Model outputs shift as underlying models are updated; data quality degrades silently; vendors change APIs without warning; and accountability for AI failures is often unclear when something goes wrong. None of those failure modes appear in a standard RAID log.

This 20-item checklist covers four AI-specific risk categories your current process is missing: model and output risks, data risks, governance and accountability gaps, and vendor dependencies. Run it before deploying a new AI system, when an existing system is updated, and at each quarterly risk review.

For every Gap identified, add a row to your AI risk register with an owner and a mitigation deadline.

Checklist Preview — Categories A & B of 4

20-Item AI Risk Assessment Checklist

Category A: Model and Output Risks — Items 1–5
1Accuracy, reliability, and confidence levels of the model are documented .
2Known failure modes have been identified — hallucination, overconfidence, edge case breakdown.
3A process exists to detect when model accuracy degrades over time (model drift).
4A human review step is in place for AI outputs with compliance, financial, or customer impact .
5You know what happens when the model receives inputs outside its training distribution .
Category B: Data Risks — Items 6–10
6The provenance of training or input data is documented — what data, from where, authorized by whom.
7A privacy and consent assessment has been completed for all data the system processes .
8Data quality risks are documented — missing data, stale data, inconsistent formats, labelling errors.
9You know what happens to data submitted to a third-party AI vendor — storage, retention, model training.
10You have a plan if data input to the system is found to be out-of-scope or improperly collected .
Gated content

Category C: Governance & Accountability (Items 11–15) · Category D: Vendor & Dependency Risks (Items 16–20) — included in the download below.

Get the Complete 20-Item Risk Assessment Checklist

Categories C (Governance and Accountability) and D (Vendor and Dependency Risks) — 10 remaining items — are in the AI Governance Starter Kit.

  • AI Governance Checklist — 25 items across 5 sections
  • AI Risk Register Template — 5 AI-specific risk categories
  • AI Decision Log Template — 5-field entry structure
  • AI Governance Framework Template — 5-section policy structure
  • AI Risk Assessment Checklist — 20 items, 4 categories

Download the AI Governance Starter Kit — Free

Free PDF  ·  Immediate download  ·  No spam

Based on the accountability framework in Authorizing the Machine — a practical guide to AI accountability for project managers. Coming soon.

Frequently Asked Questions

What risks does a standard risk assessment miss when using AI on a project?

Standard risk frameworks typically miss: model output risks (hallucination, inconsistency), data risks (provenance, poisoning, privacy), governance gaps (no named AI owner, no escalation path), and vendor dependencies (switching costs, vendor stability, SLA coverage).

How do you assess AI risk on a project?

Assess AI risk across four dimensions: model output risk (what happens if the AI is wrong?), data risk (what data is exposed and how is it handled?), governance gaps (who owns AI decisions?), and vendor dependencies (what happens if the vendor changes terms or discontinues the service?).

Start with the AI Governance Starter Kit

Five templates. One download. Free for PMs and PMO leaders responsible for AI-enabled work.

Download Free — No Spam