AI Risk Register for Research Projects

Interactive template to log AI-specific risks (bias, privacy, misuse, model drift, interpretability, regulatory) with likelihood, severity, mitigations, monitoring metrics, owner, review cadence, and examples. Saves structured entries for later export, dashboarding, and follow-up.

Interactive Tool

AI Risk Register Entry

Use this form to record one AI-related risk for a research project. Each submission captures the risk, its potential impact, a simple likelihood x severity score, mitigation and monitoring plans, and the person responsible for follow-up.

How to use: Create one entry per distinct risk; keep descriptions concise but specific. Use Likelihood and Severity scales to compute a simple risk score (Likelihood x Severity). Typical actions after saving: review register regularly, export to CSV for reporting, and link high-scoring risks to experiments, approvals, or monitoring workflows.

Example rows:

Risk descriptionPotential impactLikelihoodSeverityMitigationMonitoring metric
Training data imbalance causes subgroup performance gapsBiased conclusions; harms to underrepresented participants; reputational and regulatory riskPossible (3)Major (4)Collect additional data, apply fairness-aware reweighting, document limitationsSubgroup accuracy gap; fairness metric by cohort
Model drift after deployment in new populationIncorrect predictions, invalidated results, wasted downstream experimentsLikely (4)Moderate (3)Establish monitoring, retraining triggers; shadow testingPrediction distribution shift; validation set error
Describe the AI-related risk clearly and concisely (e.g. model bias causing unfair outcomes).
Explain consequences if the risk materializes: scientific, ethical, regulatory, safety, operational, financial, or reputational.
Estimated probability the risk will occur.
Estimated severity if the risk occurs.
Multiply Likelihood x Severity for a simple score (1-25). You may leave this blank and calculate offline or with a dashboard. Higher scores indicate priority.
Planned or implemented steps to reduce likelihood or severity. Include owners, deadlines, and acceptance criteria where possible.
Concrete metrics to detect model drift, bias, performance degradation, privacy issues, or misuse (e.g. subgroup accuracy, PSI, feature distribution stats).
Person or role responsible for the risk and for executing mitigations/reviews.
How often the owner will review this risk or when it will be re-assessed.
Current status of the risk management plan.
Optional: links to documents, datasets, experiments, model artifacts, or approvals related to this risk.
You can explore this tool now. Sign in or create an account to save your responses and return to them later.
Make this tool part of your work

Save a personal copy, bring it to your team, or tailor the questions and workflow to fit what you are hungry to improve.

Member customization and team collaboration are coming soon.

Discussion

Comments and conversation will live here.