Evaluation, model risk and AI governance.

Technologies

• Performance evaluation Task metrics, test design and reporting that show how a model performs and where it does not.
• Robustness testing Testing under noise, shift, missing inputs and edge cases.
• Bias assessment Measuring unwanted differences in performance across groups and regions.
• Adversarial testing and red-teaming Probing models for evasion, poisoning, prompt injection and data leakage.
• Uncertainty and calibration reporting Stating uncertainty with each model output so users can weigh it.
• Model risk management Inventory, tiering, independent validation and ongoing monitoring of models used in decisions.
• AI management system Organisation-wide policies, roles, controls and improvement for AI.
• AI risk management Identifying, measuring and treating AI risks across the lifecycle.
• AI impact assessment Assessing effects on people and communities before deployment and when use changes.
• Regulatory and policy alignment Mapping systems to the AI laws, codes and conventions of each market served.