What is this programme?
The AI Ethics & Governance Certificate from ecertificate.pro is an independent, self-paced online program of about 24 hours covering bias measurement, explainability, data governance, model documentation and AI audit practice. You learn to compute group fairness metrics such as demographic parity difference and the disparate impact ratio, to choose among explainability methods like SHAP, counterfactual explanations and partial dependence curves, and to assemble the technical documentation an EU AI Act or NIST AI RMF review expects. The program is issued by ecertificate.pro alone: it is not government-approved, not accredited by any national education authority, and it grants no professional licence, title or job guarantee. Each certificate carries a unique verification code that can be checked only in the ecertificate.pro database, since no third-party registry holds these records. The certificate is delivered inside your private panel on ecertificate.pro and is never sent as an e-mail attachment. It suits practitioners who already build or oversee AI systems and need a documented, repeatable governance workflow rather than a list of principles.
Who it is for
Data scientists and ML engineers who ship models, product managers who own AI features, privacy and compliance officers, internal auditors, and risk teams responsible for AI systems already in production or about to launch.
What you will be able to do
- Classify an AI system against EU AI Act risk tiers and the four NIST AI RMF functions (Govern, Map, Measure, Manage), and record in writing why each classification was chosen.
- Compute and interpret group fairness metrics - demographic parity difference, disparate impact ratio under the four-fifths rule, equal opportunity difference, false negative rate parity and calibration by group - using Fairlearn on a supplied dataset.
- Run subgroup and intersectional slice analysis to expose where aggregate accuracy hides failures for small groups, and judge when a disparity is a data collection problem rather than a model problem.
- Apply mitigation at each lifecycle point - reweighing before training, constraint-based optimisation during training, group-specific threshold tuning after training - and quantify the accuracy cost of each choice.
- Select an explainability method that fits the question being asked: SHAP for global attribution, counterfactual explanations for individual appeals, partial dependence and ICE curves for behaviour checks - and state why a saliency map is not evidence of causality.
- Write a model card and a dataset datasheet covering intended use, out-of-scope use, evaluation results by slice, known failure modes, human oversight requirements and version history.
- Assemble an audit evidence pack: data lineage records, evaluation logs, review sign-offs, red-team findings, and a post-deployment monitoring plan with drift and complaint triggers.
Modules
-
1
Risk Classification and the Governance Operating Model
-
2
Where Bias Comes From: Data, Labels and Objective Functions
-
3
Measuring Fairness: Metrics, Slices and Thresholds
-
4
Mitigation Across the Model Lifecycle
-
5
Explainability That Holds Up Under Questioning
-
6
Data Governance, Provenance and Lineage
-
7
Model Documentation: Cards, System Cards and Technical Files
-
8
Auditing, Red-Teaming and Post-Deployment Monitoring
How it works
Learning format
Self-paced written material, downloaded from your dashboard. No fixed schedule and no live sessions — you work through it when it suits you.
Material delivery
Everything appears in your dashboard as soon as your payment is confirmed. Files are served only to your signed-in session.
Prerequisites
No programming background is required for the governance, documentation and audit modules. The fairness-metric and explainability exercises are easier to follow if you can read basic Python and understand what a classifier's confusion matrix shows; a spreadsheet route is provided for every calculation if you prefer not to run notebooks.
Completion
Work through all eight modules, submit two deliverables for a supplied sample model - a completed model card and a fairness-testing report - and pass a 40-question final assessment at 70% or above. Once both parts are cleared, the certificate is issued to your private panel on ecertificate.pro.
Your certificate
Once your completion is confirmed, we prepare your certificate and upload it to your dashboard, normally within 24–48 hours. You receive an e-mail as soon as it is there — you never have to chase it.
Refunds
Because the material is delivered digitally and immediately, purchases are non-refundable once the files have been made available to your account. The full terms are in our Refund Policy.
