Cyber Wiki Topic 12
AI Security & Emerging Technology
AI system risk, governance, model protection, automation, and fast-changing technology.
Topic Directory
50 entries
AI Agents and Tool Use
An AI agent selects actions and invokes tools, services, or workflows to pursue a goal across multiple steps.
Responsible AI · AdvancedAI Bias and Fairness Assessment
AI bias and fairness assessment evaluates how data, labels, objectives, design choices, and deployment conditions can create uneven errors or harms across groups.
AI engineering assurance · AdvancedAI Configuration Reproducibility
AI configuration reproducibility means an authorized team can recreate the relevant model build or runtime behavior from recorded code, data references, parameters, dependencies, prompts, and infrastructure settings.
AI data engineering · FoundationalAI Data Preparation
AI data preparation collects, cleans, labels, transforms, and partitions information for development and evaluation.
AI operations · IntermediateAI Feedback Loops
AI feedback loops occur when a system's outputs influence the future data used to evaluate or retrain it.
AI assurance · AdvancedAI Model Artifact Audit Trail
An AI model artifact audit trail links a deployed model to its source code, training configuration, data references, parameters, evaluation results, approvals, and release package.
AI lifecycle governance · FoundationalAI Model Lifecycle
The AI model lifecycle covers conception, data, development, testing, release, operation, change, retirement, and evidence preservation.
AI product design · FoundationalAI Problem Framing
AI problem framing translates a business or public objective into a defined decision, target, user population, success measure, and set of constraints.
Governance practice · IntermediateAI Security Governance
AI security governance establishes accountable decisions, controls, evidence, and oversight across the AI lifecycle without treating AI as either ordinary software or an unknowable special case.
AI foundations · FoundationalAI System Taxonomy
An AI system taxonomy separates capabilities, learning methods, autonomy, deployment context, and intended purpose so teams use the same language when describing a system.
AI governance practice · IntermediateAI Use-Case Readiness Assessment
AI use-case readiness assessment tests whether a proposed AI capability has a defined purpose, suitable data, accountable owners, realistic benefits, acceptable failure modes, and monitorable risk.
AI security · AdvancedAdversarial Machine Learning
Adversarial machine learning studies attacks and mitigations involving model training, inference, data, outputs, and operational dependencies.
Cyber-physical risk domain · AdvancedAutomotive Cybersecurity
Automotive cybersecurity protects vehicle electronic systems, software, communications, data, update infrastructure, service tools, manufacturing, and cloud services while preserving safety, reliability, privacy, and long product lifecycles.
Decision science · AdvancedCausal Inference and Experimentation
Causal inference asks whether an intervention changed an outcome rather than merely identifying correlation.
Governance practice · AdvancedCloud AI Services Governance
Cloud AI services governance controls how managed machine-learning, generative AI, language, vision, and decision services are selected, configured, and used.
Vision AI · FoundationalComputer Vision Systems
Computer vision systems extract patterns from images or video to classify, detect, segment, track, or generate visual content.
Machine-learning governance · AdvancedData Labeling Quality and Ground Truth
Data-labeling quality and ground-truth governance define how labels are created, reviewed, reconciled, versioned, and interpreted for analytics and machine learning.
Decision science · AdvancedDecision Intelligence and Optimization
Decision intelligence combines predictions, constraints, objectives, and human judgment to choose actions rather than only estimate outcomes.
AI infrastructure · IntermediateEdge AI Deployment
Edge AI runs inference on local devices to reduce latency, support offline use, and keep some data near its source.
Embedded technology foundation · AdvancedEmbedded Systems Security
Embedded systems security protects purpose-built computing devices whose software, hardware, interfaces, timing, physical environment, and long support lifecycle are tightly coupled to a product or physical process.
Generative AI infrastructure · IntermediateEmbeddings and Vector Databases
Embeddings represent meaning as numeric vectors, and vector databases retrieve items based on similarity rather than exact wording.
Machine learning · IntermediateFeature Engineering
Feature engineering converts raw data into variables that make relevant patterns usable by a learning algorithm.
Human-centered AI · IntermediateHuman-AI Interaction Design
Human-AI interaction design shapes how people understand, question, correct, and rely on an AI system.
AI reasoning · FoundationalKnowledge Representation in AI
Knowledge representation structures facts, relationships, categories, and rules so an AI system can reason over information rather than only store it.
Cyber-physical risk domain · AdvancedMedical Device Cybersecurity
Medical device cybersecurity manages the confidentiality, integrity, availability, authenticity, resilience, and safe operation of connected or software-enabled devices throughout design, clinical use, maintenance, vulnerability response, and retirement.
Machine-learning assurance · AdvancedModel Data Leakage
Model data leakage occurs when training or evaluation uses information that would not legitimately be available at prediction time, causing performance to appear better than it will be in practice.
Machine-learning operations · AdvancedModel Drift Monitoring
Model drift monitoring detects when input distributions, relationships, outcomes, or operational conditions move away from those used to build and validate a model.
Responsible AI · AdvancedModel Explainability and Transparency
Model explainability and transparency provide appropriate information about a model’s purpose, inputs, limitations, behavior, uncertainty, and decision role to affected audiences.
Model evaluation · FoundationalModel Performance Metrics
Model performance metrics summarize different kinds of error, such as accuracy, precision, recall, calibration, and regression loss.
AI governance · AdvancedModel Risk Management
Model risk management establishes governance for selecting, approving, using, changing, challenging, and retiring models according to their impact and uncertainty.
Supply-chain security · AdvancedModel Supply-Chain Security
Model supply-chain security verifies the origin, integrity, dependencies, licenses, safety, and allowed use of models and supporting artifacts obtained or moved across organizational boundaries.
AI assurance · AdvancedModel Validation and Testing
Model validation and testing determine whether a model is appropriate, reproducible, robust, secure, fair, and effective for a defined population and decision context.
Language AI · FoundationalNatural Language Processing
Natural language processing converts human language into representations that support classification, extraction, generation, translation, and dialogue.
Deep learning · FoundationalNeural Network Fundamentals
A neural network transforms inputs through weighted layers and nonlinear functions that are adjusted during training.
Model evaluation · FoundationalOverfitting and Generalization
Overfitting occurs when a model fits training examples or noise too closely and does not generalize to new conditions.
Privacy engineering · AdvancedPrivacy-Preserving Machine Learning
Privacy-preserving machine learning applies technical and governance measures to limit exposure of personal or sensitive information during training, evaluation, inference, and sharing.
AI reasoning · IntermediateProbabilistic Reasoning in AI
Probabilistic reasoning represents uncertainty explicitly and combines evidence to estimate outcomes rather than pretending every decision is certain.
AI security concept · IntermediatePrompt Injection and Agentic AI Risk
Prompt injection attempts to influence an AI system through untrusted instructions, while agentic risk grows when a model can use tools, retrieve data, or take actions beyond producing text.
Applied AI governance · IntermediateRecommender Systems Governance
Recommender systems rank content, products, or actions using predicted relevance and feedback from prior behavior.
Machine learning · IntermediateReinforcement Learning
Reinforcement learning improves behavior through rewards and penalties received while an agent interacts with an environment.
Generative AI architecture · IntermediateRetrieval-Augmented Generation Architecture
Retrieval-augmented generation combines a generative model with selected external information to ground responses in current or controlled sources.
Autonomous systems · IntermediateRobotics and Autonomous Systems
Robotics combines sensing, AI decision logic, and physical actuation to operate in an environment.
AI reasoning · FoundationalSearch and Planning in AI
Search and planning methods explore possible states, actions, and costs to find a path toward a defined objective.
Speech and audio AI · FoundationalSpeech AI Systems
Speech AI systems recognize, synthesize, identify, or transform spoken language and other audio signals.
Machine learning · FoundationalSupervised Learning
Supervised learning trains a model from labeled examples so it can predict a class or numerical value for new observations.
AI foundations · FoundationalSymbolic and Statistical AI
Symbolic AI represents knowledge and rules explicitly, while statistical AI learns patterns from examples and probability distributions.
Predictive analytics · IntermediateTime-Series Forecasting
Time-series forecasting estimates future values from ordered observations, trends, seasonality, and related signals.
AI threat defense · AdvancedTraining Data Poisoning
Training-data poisoning manipulates examples, labels, sources, or collection processes to alter a model’s learned behavior.
Generative AI · IntermediateTransformer Architecture
Transformer models use attention mechanisms to relate tokens across a sequence and support large-scale language and multimodal systems.
Machine learning · FoundationalUnsupervised Learning and Clustering
Unsupervised learning identifies structure in unlabeled data, including clusters, latent dimensions, and unusual observations.