A–Z · Full list

All 219 terms

Every term in the Cyber & AI Governance Dictionary, alphabetically. Prefer to browse by domain?

A

AAA (AAA) Proving who you are, deciding what you can do, and logging what you did. Account lockout Temporarily disabling a login after too many failed password attempts. Accountability Having someone clearly answerable for an AI system's behavior. Advanced persistent threat (APT) A patient, well-resourced attacker that stays hidden in a network over time. Agentic AI risk The added risk when an AI doesn't just answer but takes actions on its own. AI customer An organization or individual that acquires an AI product or service. ISO/IEC 22989 AI management system (AIMS) An organization's ongoing, structured program for governing its AI use. AI objectives Specific, measurable goals the AIMS is working toward. 42001 · Clause 6 AI partner An organization supplying services to an AI producer or provider. ISO/IEC 22989 AI policy The top-level, leadership-approved statement of AI governance commitments. 42001 · Clause 5 AI producer The organization responsible for designing, building, and deploying an AI system. ISO/IEC 22989 AI provider An organization that supplies a product or service built on AI. ISO/IEC 22989 AI red teaming Deliberately attacking an AI system to find its weaknesses first. AI risk assessment Identifying and evaluating the risks a given AI system poses. 42001 · 6.1.2 AI risk treatment Deciding how to handle each identified AI risk and selecting controls. 42001 · 6.1.3 AI subject An individual affected by an AI system's decisions or outputs. ISO/IEC 22989 AI system A system that generates outputs — predictions, decisions — from defined objectives. ISO/IEC 22989 AI system impact assessment Assessing how an AI system affects people, not just the organization's own risk. 42001 · 6.1.4 AI system life cycle The full span of an AI system's existence, from design to retirement. ISO/IEC 5338:2023 AIMS scope The documented boundary of what the AI management system covers. 42001 · Clause 4 Annex A controls A reference set of AI-specific controls selected based on actual risk. 42001 · Annex A Annex A.10 — Third-party and customer relationships Controls for AI risk from suppliers and expectations with customers. Annex A.10 Annex A.2 — Policies related to AI Controls for establishing and reviewing an organization's AI policies. Annex A.2 Annex A.3 — Internal organization Controls for AI roles, reporting lines, and ethics escalation. Annex A.3 Annex A.4 — Resources for AI systems Controls covering the data, tooling, compute, and people AI depends on. Annex A.4 Annex A.5 — Assessing impacts of AI systems Controls for actually running AI system impact assessments. Annex A.5 Annex A.6 — AI system life cycle Controls spanning design through decommissioning of an AI system. Annex A.6 Annex A.7 — Data for AI systems Controls for the quality, provenance, and prep of data feeding AI systems. Annex A.7 Annex A.8 — Information for interested parties Controls for what's documented and disclosed about an AI system. Annex A.8 Annex A.9 — Use of AI systems Controls for intended use, objectives, and monitoring once deployed. Annex A.9 Annex B — Implementation guidance Informative guidance for actually implementing Annex A's controls. 42001 · Annex B Annex C — AI-related objectives and risk sources A catalogue of typical AI risk sources and related objectives. 42001 · Annex C Annex D — Use across domains and sectors Guidance on applying 42001 across sectors and alongside other standards. 42001 · Annex D Assess Step 5 of the RMF — an independent check on whether controls actually work. NIST RMF · Step 5 of 7 Asset Anything worth protecting — data, a system, a device, or a reputation. Attack surface Every point an attacker could try to get in through. Authorize Step 6 of the RMF — a senior official formally approves the system to run. NIST RMF · Step 6 of 7 Awareness Making sure staff understand the AI policy and their part in it. 42001 · Clause 7

C

Categorize Step 2 of the RMF — classify the system by how much a loss would hurt. NIST RMF · Step 2 of 7 Certification audit The independent audit confirming an AIMS actually meets ISO/IEC 42001. Certification body The accredited organization authorized to audit and certify an AIMS. 42001 · certification CIA triad (CIA) The three goals of security: keep data private, unaltered, and available when needed. Cloud IAM roles Predefined permission sets assigned to a user or service in the cloud. Cloud security posture management (CSPM) Tools that continuously scan cloud environments for misconfigurations. Communication Defining what AIMS information gets shared, with whom, and how. 42001 · Clause 7 Competence Making sure people doing AIMS work actually have the skills to do it. 42001 · Clause 7 Conditional access Access rules that adapt to context — location, device health, risk. Context of the organization Identifying the issues and interested parties relevant to the AIMS. 42001 · Clause 4 Continual improvement Keeping the AIMS getting better over time instead of standing still. 42001 · Clause 10 Control (preventive, detective, corrective) A safeguard that reduces risk — preventive, detective, or corrective. Control objective The specific goal a given Annex A control exists to achieve. 42001 · Annex A Corrective action Fixing the root cause of a nonconformity, not just the symptom. 42001 · Clause 10 Credential stuffing Trying leaked username/password pairs against other sites, betting on reuse. Cross-site scripting (XSS) Injecting malicious script into a site so it runs in another visitor's browser. Cyber insurance Insurance covering breach response, legal fees, and lost income from an incident.

D

Data classification Labeling data by sensitivity so the right controls apply automatically. Data loss prevention (DLP) Tools that detect and block sensitive data from leaving improperly. Data minimization Collecting and keeping only the data actually needed for the task. Data poisoning Deliberately corrupting training data so a model learns the wrong behavior. Data provenance The documented origin and history of a dataset. Data quality How accurate, complete, and representative the data feeding AI actually is. Data residency Where data is physically stored, which can determine which laws apply. Data retention How long an organization keeps data before deleting it. Deepfake fraud Using AI-generated audio or video to convincingly impersonate someone. Defense in depth Layer multiple controls so one failure doesn't leave you fully exposed. Demilitarized zone (DMZ) A network zone between the public internet and the internal network. Detect The CSF function for spotting attacks and compromises as they happen. NIST CSF 2.0 · Detect Digital certificate An electronic credential proving a website or signer is who they claim. Digital forensics A structured investigation after an incident to determine what happened. Disaster recovery plan (DRP) The technical plan for restoring IT systems after a major disruption. Distributed denial-of-service (DDoS) Flooding a system with traffic until it can't serve real users. DNS filtering Blocking lookups for known-malicious domains before a device can connect. Documented information The records an AIMS has to keep as evidence it's actually operating. 42001 · Clause 7 Due care vs. due diligence Research before acting, versus ongoing effort to maintain protection after.

I

IaaS, PaaS & SaaS Three layers of cloud service: raw infrastructure, a platform, or a finished app. Identify The CSF function for cataloguing assets, data, and risk before protecting them. NIST CSF 2.0 · Identify Identity and access management (IAM) The systems that create accounts, verify identity, and control access. Identity provider (IdP) The central service that verifies identity and issues trusted proof of it. Immutable backup A backup that can't be altered or deleted, even by an administrator. Implement Step 4 of the RMF — actually put the chosen controls in place. NIST RMF · Step 4 of 7 Incident response plan (IRP) A written, rehearsed plan for the first hours and days of an incident. Indicator of compromise (IOC) Evidence a system has already been breached. Information security management system (ISMS) An ongoing program for managing information security, built on ISO/IEC 27001. Insider threat Harm caused by someone with legitimate access, malicious or careless. Interested parties Everyone with a stake in how an organization manages its AI. 42001 · Clause 4 Internal audit A self-check confirming the AIMS meets the standard and its own requirements. 42001 · Clause 9 Intrusion detection & prevention systems (IDS/IPS) Watches traffic for attack patterns — one alerts, the other can block automatically. ISO/IEC 22989 Defines the shared AI vocabulary the rest of the AI standards family builds on. ISO/IEC 22989:2022 ISO/IEC 23053 A framework describing the building blocks of machine-learning-based AI systems. ISO/IEC 23053 ISO/IEC 23894 Adapts general risk management principles specifically to AI. ISO/IEC 23894:2023 ISO/IEC 27001 The international standard for building and certifying an ISMS. ISO/IEC 38507 Gives governing bodies guidance on overseeing an organization's use of AI. ISO/IEC 38507:2022 ISO/IEC 42001 The international standard for building and certifying an AI management system. ISO/IEC 42001:2023 ISO/IEC 42005 Detailed guidance for conducting an AI system impact assessment. ISO/IEC 42005:2025 ISO/IEC 42006 Sets the requirements certification bodies must meet to audit against 42001. ISO/IEC 42006:2025 ISO/IEC 5338 Defines the processes that make up an AI system's life cycle. ISO/IEC 5338:2023 ISO/IEC TR 24027 A technical report on measuring bias in AI systems and AI-aided decisions. ISO/IEC TR 24027:2021

M

Machine learning (ML) Building AI systems that learn patterns from data instead of fixed rules. Malware Software built to damage, disrupt, or gain unauthorized access to a system. Man-in-the-middle attack (MITM) An attacker secretly intercepts communication between two parties. Manage The AI RMF function for treating AI risk and responding when something goes wrong. NIST AI RMF · Manage Managed detection and response (MDR) An outsourced service that actively monitors and responds to threats for you. Management review Leadership's periodic review of how the AIMS is performing. 42001 · Clause 9 Map The AI RMF function for understanding a specific AI system's context and risks. NIST AI RMF · Map Measure The AI RMF function for actually measuring and tracking identified AI risk. NIST AI RMF · Measure Membership inference Determining whether someone's data was used to train a model. Misconfiguration A security setting left insecure — a common cause of cloud data exposure. MITRE ATLAS A knowledge base of real-world attacker tactics specifically against AI systems. Model extraction Querying a model repeatedly to steal a close copy of it. Model inversion Reconstructing sensitive training data by studying a model's outputs. Monitor Step 7 of the RMF — ongoing tracking of security after the system is live. NIST RMF · Step 7 of 7 Monitoring and measurement Tracking whether the AIMS and its controls are actually working. 42001 · Clause 9 Multi-factor authentication (MFA) Requiring two or more proofs of identity to log in, not just a password.

P

Passkeys A passwordless sign-in method tied to a device that resists phishing. Password manager Generates and stores a unique, strong password for every account. Patch management Finding, testing, and installing updates that fix known vulnerabilities. Payment Card Industry Data Security Standard (PCI DSS) Security requirements for any organization handling payment card data. Penetration testing An authorized simulated attack to find exploitable weaknesses. Personally identifiable information (PII) Data that can identify a specific individual on its own or combined. Phishing A fake message designed to trick someone into handing over credentials or money. Plan-Do-Check-Act (PDCA) The Plan-Do-Check-Act cycle that Clauses 4–10 of every ISO management standard follow. ISO Harmonized Structure Planning of changes Planning changes to the AIMS in a controlled way as AI use evolves. 42001 · Clause 6 Policy, standard & procedure A policy sets the rule, a standard sets the requirement, a procedure the steps. Prepare Step 1 of the RMF — set priorities and groundwork before anything else. NIST RMF · Step 1 of 7 Private / self-hosted AI Running an AI model on infrastructure you control, so data never leaves. Private / sovereign cloud Infrastructure built to keep data under an organization's or country's control. Privileged access management (PAM) Extra controls around the small number of highly powerful admin accounts. Prompt injection Hidden or direct instructions that trick an AI into doing something unintended. Protect The CSF function for putting safeguards in place to manage known risk. NIST CSF 2.0 · Protect Protected health information (PHI) Individually identifiable health information covered under HIPAA. Public key infrastructure (PKI) The system of certificates and keys that proves identity without a shared secret.

S

Salting Adding random data before hashing so identical passwords don't match. Security Assertion Markup Language (SAML) An older standard for passing proof of login between systems for SSO. Security awareness training Ongoing education that teaches staff to recognize phishing and handle data. Security information and event management (SIEM) Collects logs across systems and correlates them to surface suspicious activity. Security operations center (SOC) The team that watches security alerts and responds around the clock. Security posture An organization's overall readiness against cyber risk right now. Select Step 3 of the RMF — choose the controls that fit the system's risk level. NIST RMF · Step 3 of 7 Separation of duties Splitting a sensitive task across more than one person to prevent abuse. Session hijacking Stealing a valid login session to act as an already-authenticated user. Shadow AI AI tools employees use on their own, without IT's knowledge or approval. Shared responsibility model The split between what the cloud provider secures and what you secure. Single sign-on (SSO) Logging in once to access many connected applications. Smishing & vishing Phishing delivered by text message or phone call instead of email. SOC 2 An auditor's report on how well a service provider's security controls hold up. Social engineering Manipulating a person, not a system, into breaking security procedure. Spear phishing Phishing aimed at one specific person using details that make it believable. SQL injection (SQLi) Sneaking database commands into a web form to run them against the database. Statement of Applicability (SoA) The document listing which Annex A controls apply, and why. 42001 · 6.1.3 / Annex A Supply chain attack Compromising a trusted vendor to reach their downstream customers.

Sources & disclaimer

  • Definitions on this page are original, plain-language explanations.
  • The ISO standards themselves are available for purchase from iso.org.
  • This page is general information, not legal or certification advice.