Accountability (AI)
- Definition
- Accountability in AI means that a named person or organization is answerable for an AI system's decisions and effects, with the authority and the records needed to act on that responsibility.
Showing 20 of 74 terms
AI Acceptable Use Policy
- Definition
- An AI acceptable use policy is the written rule set that tells staff which AI tools they may use, with which data, for which tasks, and who approves exceptions.
AI Alignment
- Definition
- AI alignment is the effort to make an AI system's goals and behavior match what its operators and users actually intend, including their values and constraints, rather than a literal or convenient…
AI Audit
- Definition
- An AI audit is a structured, evidence-based review of an AI system that checks whether it does what was claimed, follows policy and law, and has controls that actually work.
AI Audit Log
- Definition
- An AI audit log is a tamper-resistant record of every action an AI system took, including prompts, model responses, tool calls, data accessed, and human approvals, kept so the activity can be…
AI Component of a System
- Definition
- In SOC 1 and SOC 2 reporting, the AI component of a system is the part of a service organization's system that uses AI, including its models, data, configuration, and the controls…
AI Content Watermarking
- Definition
- AI content watermarking embeds a hidden, machine-readable signal in generated text, images, audio, or video so that the output can later be identified as AI-made.
AI Governance
- Definition
- AI governance is the set of policies, roles, controls, and records a company uses to decide which AI systems it runs, how they are built or bought, and who answers for their…
AI in Healthcare
- Definition
- AI in healthcare is the use of machine learning and language models in clinical, administrative, and revenue cycle work, under privacy rules that govern where patient data may be processed.
AI Incident
- Definition
- An AI incident is an event in which an AI system's use or malfunction causes harm, such as an unfair decision about a person, a leak of confidential data, or an unauthorized…
AI Inventory
- Definition
- An AI inventory is a maintained register of every AI system a company builds, buys, or uses, recording each one's owner, purpose, data, connected systems, and risk tier.
AI Literacy
- Definition
- AI literacy is the knowledge and judgment people need to use AI systems well in their role. The EU AI Act requires providers and deployers to take measures that support it among…
AI Risk Classification (EU AI Act)
- Definition
- AI risk classification under the EU AI Act is the process of assigning an AI system to one of four tiers, prohibited, high, limited, or minimal risk, which determines the legal duties…
AI Risk Management
- Definition
- AI risk management is the ongoing process of identifying what could go wrong with an AI system, judging how likely and how serious each harm is, and deciding what to do about…
AI Safety
- Definition
- AI safety is the practice of preventing AI systems from causing harm, whether through errors, misuse, unintended behavior, or loss of human control, using testing, constraints, monitoring, and oversight throughout the system's…
AI Security
- Definition
- AI security is the practice of protecting AI systems, including models, data, prompts, and agents, from attack and misuse, and of preventing a manipulated AI system from harming the business that runs…
AI System
- Definition
- An AI system is a deployed product made of one or more AI models plus the interface, data pipelines, integrations, and controls around them. Regulation such as the EU AI Act applies…
Approval Workflow
- Definition
- An approval workflow is a control that pauses an AI agent before a consequential action, presents the proposed change to a named person, and proceeds only when that person approves it.
Auditable AI
- Definition
- Auditable AI is an AI system built so that an independent reviewer can reconstruct what it did, why, with which data and model version, and who approved the result. It rests on…
Automated Decision-Making
- Definition
- Automated decision-making is the use of software to reach decisions about people without meaningful human involvement. GDPR Article 22 restricts such decisions when they have legal or similarly significant effects on a…
Automation Bias
- Definition
- Automation bias is the human tendency to accept an automated system's output without checking it, even when the output is wrong and the person has the evidence to catch it.
Bias (AI)
- Definition
- Bias in AI is a systematic error in a model's outputs. It can be statistical, a consistent gap between predictions and reality, or social, a pattern that treats groups of people unequally.
Bias Detection and Fairness Auditing
- Definition
- Bias detection and fairness auditing are the methods used to find out whether a model treats groups of people differently and to document whether those differences are justified.
Conformity Assessment
- Definition
- A conformity assessment is the procedure by which the provider of a high-risk AI system demonstrates, before placing it on the EU market, that the system meets the requirements of the EU…
Content Moderation
- Definition
- Content moderation is the review of user-generated or AI-generated content to decide whether it is allowed, using classifiers, written policies, and human reviewers for the cases a model cannot settle.
Content Provenance
- Definition
- Content provenance is a verifiable record of where a piece of digital content came from and how it was edited, attached to the file as signed metadata.
Counterfactual Explanation
- Definition
- A counterfactual explanation states the smallest change to a model's inputs that would have produced a different outcome, for example what would have turned a declined application into an approved one.
Data Governance
- Definition
- Data governance is the set of roles, policies, and processes that decides who owns each kind of data, who may use it and how, and how its quality and security are maintained.
Data Lineage
- Definition
- Data lineage is the documented path data takes from its original source through every transformation, system, and calculation to the report, dashboard, or AI model that uses it.
Data Sovereignty
- Definition
- Data sovereignty is the principle that data is subject to the laws and governance of the country where it is collected or stored, and the practice of keeping control of it accordingly.
Datasheets for Datasets
- Definition
- Datasheets for datasets are structured documents that record why a dataset was built, what it contains, how it was collected and labeled, and which uses it does and does not support.
Decision Traceability
- Definition
- Decision traceability is the ability to reconstruct, after the fact, how an AI system reached a specific decision: the inputs it saw, the rules or model it applied, the steps it took,…
Deepfake
- Definition
- A deepfake is synthetic audio, video, or imagery generated by AI to depict a real person saying or doing something they never did, usually convincingly enough to deceive a viewer.
DORA (Digital Operational Resilience Act)
- Definition
- DORA, the Digital Operational Resilience Act, is the EU regulation that requires financial entities to manage ICT risk, report incidents, test resilience, and control the third-party technology providers they depend on, including…
DPIA and FRIA
- Definition
- A DPIA is the GDPR's data protection impact assessment for high-risk processing of personal data, and a FRIA is the EU AI Act's fundamental rights impact assessment that certain deployers of high-risk…
Ethical AI
- Definition
- Ethical AI is the practice of designing, deploying, and operating AI systems so that they respect people's rights, treat them fairly, and remain under meaningful human accountability. It turns stated values into…
EU AI Act
- Definition
- The EU AI Act is the European Union's regulation of artificial intelligence, sorting AI systems by risk and imposing duties on the companies that build or use them, inside and outside Europe.
Explainable AI (XAI)
- Definition
- Explainable AI (XAI) is the set of methods that make an AI model's outputs understandable to people, by showing which inputs drove a result and how the result would change if they…
Facial Recognition
- Definition
- Facial recognition is a biometric technology that identifies or verifies a person by comparing the geometry of their face with stored templates. It is tightly regulated and carries high privacy and bias…
Fairness (AI)
- Definition
- Fairness in AI is the property that a model's predictions or decisions do not systematically disadvantage people because of characteristics such as race, sex, age, or disability. It is measured with specific,…
Feature Importance
- Definition
- Feature importance is a measure of how much each input variable contributes to a model's predictions. Teams use it to check a model's logic, catch information leakage, and explain the model to…
GDPR (and AI)
- Definition
- The GDPR is the European Union's data protection law, and it governs any AI system that trains on or processes personal data of people in the EU, including automated decisions about them.
General-Purpose AI (GPAI) Model
- Definition
- A general-purpose AI (GPAI) model is the EU AI Act's legal term for a model, such as a large language model, that performs many distinct tasks and can be built into many…
Governance Plane
- Definition
- A governance plane is the layer of an AI platform that enforces identity, permissions, policy, and audit logging on every model call and agent action, separate from the models themselves.
Guardrails
- Definition
- Guardrails are the controls placed around a generative AI model that check inputs and outputs against rules so the system stays within its intended scope, tone, and legal limits.
High-Risk AI System
- Definition
- A high-risk AI system is one the EU AI Act places in its most regulated permitted tier because it affects safety or fundamental rights, triggering strict duties for its provider and its…
HIPAA (and AI)
- Definition
- HIPAA, the Health Insurance Portability and Accountability Act, is the US law that sets privacy and security rules for patient health information, including the information AI systems process for providers, health plans,…
Human Oversight
- Definition
- Human oversight is the set of design features and assigned people that let an organization monitor an AI system, interpret its output, and override or stop it. The EU AI Act requires…
Human-in-the-Loop
- Definition
- Human-in-the-loop is a design in which a person reviews, corrects, or approves an AI system's output before it takes effect, so that a human decision stands between the model and the outcome.
Interpretability
- Definition
- Interpretability is the degree to which a person can understand how a model turns inputs into outputs by examining the model itself, such as its rules, weights, or structure.
ISO/IEC 42001
- Definition
- ISO/IEC 42001 is the international standard for an AI management system, a certifiable set of policies, roles, and controls for developing and using AI responsibly across an organization.
Levels of Agent Autonomy
- Definition
- Levels of agent autonomy are a graded scale of how much an AI agent may decide and do without a person, from only suggesting actions to acting alone and reporting afterward.
LIME (Local Interpretable Model-Agnostic Explanations)
- Definition
- LIME (Local Interpretable Model-Agnostic Explanations) is a technique that explains one prediction from any machine learning model by testing small variations of the input and fitting a simple model to the results.
LLM Observability
- Definition
- LLM observability is the practice of recording and analyzing what a language model application does in production, including prompts, responses, tool calls, latency, token usage, errors, and quality signals.
Model Card
- Definition
- A model card is a structured document published with an AI model that describes its intended uses, training data, evaluation results, limitations, and license, so buyers and auditors can judge its fitness.
Model Evaluation
- Definition
- Model evaluation is the measurement of how well a model performs against defined criteria, using held-out data, task-specific tests, and safety checks, before and after it is put into service.
Model Risk Management
- Definition
- Model risk management is the discipline of inventorying, validating, and controlling the models a company relies on for decisions, developed in banking under SR 11-7 and now applied to AI systems generally.
Model Validation
- Definition
- Model validation is the independent review of a model's design, data, performance, and limitations to confirm it is fit for its intended use before and during deployment.
NIST AI Risk Management Framework
- Definition
- The NIST AI Risk Management Framework is a voluntary US guide for managing AI risk through four functions, Govern, Map, Measure, and Manage, that companies adapt to their own systems and sector.
Personally Identifiable Information (PII)
- Definition
- Personally identifiable information (PII) is any data that can identify a specific person on its own or combined with other data, such as a name, address, account number, or biometric record.
Post-Market Monitoring
- Definition
- Post-market monitoring is the ongoing collection and review of evidence about how an AI system performs once it is in real use. The EU AI Act requires it of every provider of…
Provider vs Deployer
- Definition
- Under the EU AI Act, a provider develops an AI system and places it on the market under its own name, while a deployer uses the system in its operations, and each…
Reporting Bias
- Definition
- Reporting bias occurs when the frequency of events, outcomes, or opinions in a dataset differs from their frequency in reality because people record some things more readily than others.
Responsible AI
- Definition
- Responsible AI is the practice of developing and operating AI systems so that they are fair, transparent, accountable, safe, and respectful of privacy, backed by concrete policies, controls, and review processes rather…
Risk Management System (EU AI Act)
- Definition
- A risk management system under the EU AI Act is the documented, continuous process a provider of a high-risk AI system must run to identify, evaluate, and reduce risks across the system's…
Selection Bias
- Definition
- Selection bias arises when the data used to train or evaluate a model is not representative of the population the model will serve, because of how the records were chosen.
Shadow AI
- Definition
- Shadow AI is the use of AI tools, models, or features at work without the knowledge or approval of IT, security, or governance owners. It most often starts with company data pasted…
Shapley Values (SHAP)
- Definition
- Shapley values are a game theory method that splits a single model prediction among its input features, showing how much each one pushed the result up or down. SHAP is the most…
SOC 2 (and AI)
- Definition
- SOC 2 is an attestation report in which an independent CPA firm tests a service provider's controls over customer data. It covers a vendor's AI features only when they sit inside the…
Sovereign AI
- Definition
- Sovereign AI is AI capability that a nation or organization controls within its own legal jurisdiction, covering the models, the data, the infrastructure, and the rules that govern them.
Tool Governance
- Definition
- Tool governance is the set of policies and controls that determine which tools an AI agent may call, with what data, under whose identity, and with what approval or logging requirements.
Transparency (AI)
- Definition
- Transparency in AI is the practice of telling people, in understandable terms, when AI is being used, what it does, what data it relies on, and what its limits are.
Transparency Obligations
- Definition
- Transparency obligations are the EU AI Act's duties to tell people when they are interacting with an AI system, to mark AI-generated content, and to give deployers of high-risk systems the information…
Trustworthy AI
- Definition
- Trustworthy AI is an AI system that is valid, reliable, safe, secure, accountable, transparent, explainable, privacy-enhanced, and fair, as defined by the NIST AI Risk Management Framework. Each property is something a…
No terms match your search.
Try a broader keyword or another category.