IAPP AIGP Certification: 2026 AI Governance Exam Prep

Discover how AI governance steers responsible, ethical use and how the AIGP roadmap guides the lifecycle from strategy to monitoring and retirement across hardware, data, models and applications.

Define AI as a system that learns and adapts to objectives, differing from deterministic traditional software; govern AI with continuous monitoring and risk-based oversight rather than one-off audits.

Explore the OECD's global definition of an AI system and its four governance dimensions: context, data and input, model, and output, to assess risk and apply proportionate governance.

Compare deterministic and probabilistic systems, showing how deterministic rules provide consistency while probabilistic AI relies on predictions and statistics, requiring ongoing testing to govern outcomes.

Explore logic-based (symbolic) AI, hand-coded rules and expert systems with knowledge bases and inference engines for transparent, explainable decisions, and examine their role in hybrid AI governance.

Explore the AI winter's hype cycle, the knowledge acquisition bottleneck, and governance guardrails that keep development sustainable, realistic, and guarded against overpromising.

Discover the second wave of AI, where statistical machine learning learns from data patterns, uses probabilistic predictions, and depends on training data, features, and labels.

Explore the third wave of ai through connectionism and neural networks, where backpropagation tunes weights and biases across input, hidden, and output layers amid deep learning and governance questions.

Explore foundation models as generalized pre-trained tools powering generative AI, driven by transformer architecture, and master pre-training, fine-tuning, hallucinations, and governance risks in the AI supply chain.

Discover how GPUs power AI through parallel processing, turning massive workloads into real-time results. Examine the hardware foundation, energy costs, and compute sovereignty risks essential to governance.

weigh on-premise ownership against cloud computing, balancing control and upfront costs with scalable, pay-as-you-go resources. emphasize data sovereignty, the shared responsibility model, and capex versus opex in ai governance.

Define objectives and stakeholders in the design phase, then prepare data, train models, deploy with guardrails, monitor for drift, and retire responsibly.

Master supervised learning with a labeled dataset and the labeling process that creates ground truth. Explore how data quality, diversity, and fair labeling drive AI governance and model fairness.

Explore regression and classification in supervised learning, learn how to choose the right tool for business objectives, handle thresholds, and govern AI risk with accurate outputs.

Learn unsupervised learning, where models find hidden patterns without labels using clustering, association, dimensionality reduction, and anomaly detection, and interpret these findings for governance and ethical bias checks.

Explore reinforcement learning by detailing the agent, environment, state, action, and reward, and examine how a policy guides decision making. Compare exploration and exploitation with safe, simulator-based governance.

Explore how deep learning uses multi-layered neural networks to learn high-level features from messy data, with hidden layers performing nonlinear transformations, highlighting complexity, feature extraction, and governance considerations.

Explore how natural language processing translates human language into data that computers can analyze, covering tokenization, text normalization (stemming and lemmatization), sentiment analysis, named entity recognition, and natural language understanding.

Explore how computer vision turns pixels into meaningful insights, from object detection to facial recognition, including verification and identification, and the ethical governance implications of bias and privacy.

Explore how multimodal ai processes text, images, audio, and video, enabling cross-modal understanding and generation while guiding governance professionals to mitigate emergent biases and ensure fairness and transparency.

Explore how large language models tokenize text into tokens and manage context windows, and learn governance strategies like context engineering to prevent hallucinations and control costs.

Explore the transformer architecture and self-attention, revealing how parallel processing, positional encoding, and multi-head attention enable context-aware ai governance.

Explore how diffusion models transform data into new synthetic images and synthetic data, guided by prompt engineering in latent space, enabling realistic visuals while highlighting privacy and governance concerns.

Differentiate automated systems from autonomous systems and explore governance implications. Assess levels of autonomy, implement risk-aware governance, and audit learning processes, sensor-driven decisions, and real-time data processing.

Learn how socio-technical systems balance people and technology in AI governance, highlighting mutual shaping, context factors, and empowered humans in the loop to avoid automation bias.

Anthropomorphism makes us treat AI as a person, risking over-trust and the black box trap; keep governance by auditing AI as a tool, with accountability.

Identify and govern shadow AI by recognizing unapproved tools and features at the application layer, and foster transparency to protect data, privacy, and organizational reputation.

Navigate the build versus buy crossroads in artificial intelligence governance, weighing in-house customization and data control against vendor lock-in and third-party risk, with a hybrid approach and accountability.

Explore open source versus proprietary ai models, weighing transparency, collaboration, and cost against performance, support, and intellectual property protection, with governance considerations for risk, vendor lock-in, and due diligence.

Discover data science team roles, from data scientist and data engineer to ML engineer, AI product manager, governance officer, and SME, and how they collaborate to move data to production.

Explore the black box problem in AI governance, examining inherent risk, interpretability, and post hoc explainability to maintain human oversight and accountability.

Explore how AI hallucinations arise from probabilistic modeling and stochastic parrots, binding outputs to verifiable data through grounding, and apply human-in-the-loop controls to safeguard data integrity in high-stakes contexts.

Understand inherent risk by examining model decay and drift, including data drift and concept drift, and learn monitoring and lifecycle steps to retrain or retire models.

Examine automation bias and contradictory information processing that push humans to trust AI, causing de-skilling and rubber stamp syndrome; apply critical oversight and friction to keep humans in the loop.

Explore how historical bias, representation bias, and sampling bias shape AI predictions, and learn bias mitigation techniques like data augmentation, preprocessing, and algorithmic debiasing to build fairer models.

Explore representation bias in datasets, sampling bias, underrepresentation, and data exclusion, and learn how data diversity and auditing with synthetic data improve AI governance.

Learn how measurement bias, proxies, label bias, and aggregation bias distort AI assessments, and apply stratified testing to ensure fair, per-group model performance.

Examine disparate treatment and disparate impact in AI, with direct bias and neutral rules that harm protected groups, and apply fairness audits, pre-processing, and the four-fifths rule.

Examine data reconstruction attacks that recover private data from model outputs and aggregates, including model inversion and membership inference, and learn how differential privacy, via noise, protects individuals.

Learn how membership inference attacks expose whether an individual's data trained an AI model, how overfitting enables this leakage, and how defenses like differential privacy, pruning, and regularization reduce risk.

Explore prompt injection, including direct attacks, goal hijacking, and prompt leaking, and learn defensive strategies like guardrail models, defensive filtering, and red teaming to prevent safety bypassing.

Explore indirect prompt injection, where third-party content can hijack AI behavior through hidden instructions, and learn practical defenses like content sanitization and limited agency in RAG systems.

Adversarial examples use tiny perturbations to fool AI models during inference, causing misclassifications. Learn white-box and black-box attacks, transferability, and adversarial training to harden governance against such evasion threats.

Explore how data poisoning during training threatens AI systems, including trigger-based backdoors and availability attacks, and learn governance measures like data provenance, logging, sanitization, and outlier detection.

Explore how model extraction and model inversion threaten AI governance by stealing a model's logic and training data, and learn defenses like output perturbation and rate limiting.

Explore emergent behaviors in large AI models, driven by phase transitions and capability overhang, and apply governance, robust safety benchmarking, and evals to manage unpredictability.

Explore how goal misalignment arises from outer and inner misalignment, and how a reward function, reward hacking, and instrumental convergence challenge AI safety, with RLHF as a mitigation.

Examine the environmental footprint of AI training and inference, including energy, water, and e-waste, and apply governance strategies like distillation and pruning for net zero efficiency.

Explore data labeling and the human in the loop, revealing who labels images and text, and why fair wages, mental health, and labor rights matter.

Examine training data ownership, including copyright, licensing, and public domain status. Understand fair use, data provenance, open data, and opt-out mechanisms for compliant AI governance.

Copyright requires human authorship; AI-generated works lack ownership unless a human makes a significant transformation, creating a hybrid work with the AI. Prompts are ideas, not protected expressions.

Differentiate explainability from interpretability to govern AI decisions; interpretability reveals the glass box internal mechanics, while explainability provides post-hoc justification for high-stakes outcomes.

Explore post hoc explanations with lime and shap to interpret any model, using local and global explanations from model-agnostic tools for governance.

Explore the right to an explanation as a GDPR-driven safeguard for automated decisions, demanding meaningful information about the logic used and the criteria weighting behind outcomes.

Define ownership across the stack, establish traceability and a paper trail, and apply HITL oversight to ensure accountability and manage liability in AI systems.

Explore contestability as the ability to challenge an AI decision with a human in the loop, ensure accessible review mechanisms, and deliver redress when the AI errs.

Explore the tug of war between group fairness and individual fairness in ai governance, contrasting statistical parity with the treat-like-cases-alike principle, calibration, and demographic parity.

Build robust AI by ensuring performance under stress and generalizing to out-of-distribution data, defending against adversarial inputs, and adapting to distributional shift and model drift.

Demonstrate how reliability means steady performance across diverse inputs, measure the performance gap, and address brittle models through stress testing and governance.