This glossary is for security leaders and practitioners at companies of 100 to 2,500 people, where a small team handles scanning, audits, vendor reviews and board reporting at once. Each entry gives a plain definition and why the term matters to a lean team, with a link to the article that goes deeper where one exists.
A to C
AI agent (for security work)
An AI agent for security work is software that takes a goal, such as getting a finding fixed, and carries it across your tools over days: reading the scanner, finding the owner, drafting the ticket, following up and checking the result. For a small team, the value is follow-through on coordination work, with a person approving anything that changes a system or gets sent.
Read more: AI Agents for Security Teams: What They Do and Where They Fit
Alert triage
Alert triage is sorting incoming security alerts to decide which are real, how urgent each one is and who acts next. The first pass usually comes from the team, the SIEM or EDR rules or an MDR provider. Triage ends in a decision, but an alert that exposes a weakness still needs an owner, a ticket and a checked fix, or it comes back.
Read more: Security Alert Triage for Small Teams: From Queue to Resolution
Bring your own model key
Bring your own model key, or BYOK, means an AI product calls the language model through your own account with the model provider. Your existing data terms, regional settings and spend controls then apply to what the model sees. The vendor stays in the data path, so still ask what it stores and logs. In cloud security, BYOK more often means supplying your own encryption keys, so check which meaning a vendor intends.
Read more: 10 Questions to Ask AI Security Vendors Before Buying
CAIQ
The CAIQ is the Consensus Assessments Initiative Questionnaire from the Cloud Security Alliance, a standard set of yes-or-no questions about a cloud provider's controls. Its questions map to the CSA's Cloud Controls Matrix. Completing it once and keeping it current gives a small team a ready answer set for buyers who accept a standard form.
Read more: How to Answer Security Questionnaires Faster (2026 Guide) and Vendor Security Assessment: A Process a Lean Team Can Run
CIS Benchmarks
The CIS Benchmarks are secure configuration guides from the Center for Internet Security, built by consensus among security practitioners, with guides for cloud providers, operating systems, network devices and server software. Each recommendation names a setting, the value to use and how to check it. Cloud security tools often test against them, so they give a small team a shared baseline for what counts as a misconfiguration.
Read more: Cloud Misconfiguration Remediation: Fix It at the Source
CISA KEV catalog
The KEV catalog is CISA's public list of Known Exploited Vulnerabilities: CVEs with evidence of exploitation in the wild. On June 10, 2026, BOD 26-04 replaced BOD 22-01 as the directive that sets federal remediation deadlines, with KEV status as one of its four factors. It binds only US federal civilian agencies. For everyone else, the catalog is free evidence of what to fix first.
Read more: CVSS vs EPSS vs KEV: How to Prioritize What Your Team Can Fix and Vulnerability Management for a Small Security Team (2026 Guide)
Compliance automation platform (GRC platform)
A compliance automation platform connects to your cloud, identity, code, HR and device tools, maps their settings to framework controls such as SOC 2, tests those controls and stores audit evidence. Vanta and Drata are examples; many people call them GRC platforms, for governance, risk and compliance. The platform is the system of record; people still fix the failures it finds and approve the evidence.
Read more: Is Vanta Enough for SOC 2? What It Automates and What's Left
Continuous control monitoring
Continuous control monitoring is testing controls on a schedule against live system data, instead of once before an audit. If MFA is switched off for a user or a storage bucket loses encryption, the test fails and the owner hears about it. A small team then fixes problems in the week they appear instead of the month before an audit.
Read more: Is Vanta Enough for SOC 2? What It Automates and What's Left
Control drift
Control drift is when a control that once worked quietly stops working, usually through an ordinary change nobody flagged: MFA missing on a new admin account, or logging off on a new storage bucket. Program work drifts too, as tickets go quiet and owners change teams. A short weekly check for drift signals catches both before an auditor does.
Read more: Security Program Management Without a PMO: A Lean Guide and Cloud Misconfiguration Remediation: Fix It at the Source
Copilot
A copilot is an AI assistant that works inside one product, such as a SIEM or code editor, and helps while you use it: writing a query, explaining an alert. It waits for your next prompt. An AI agent, by contrast, carries a task across several tools after you hand it off. Many teams use both.
Read more: AI Agents for Security Teams: What They Do and Where They Fit
CSPM
CSPM, or cloud security posture management, is software that keeps reading the settings in your cloud accounts and flags any that break a baseline, such as public storage, open network ports or logging switched off. It finds and ranks misconfigurations; it does not decide who fixes them. Each finding still needs an owner, a fix at the source and a check in the cloud that the setting changed.
Read more: Cloud Misconfiguration Remediation: Fix It at the Source
CVE
A CVE is the public ID given to one known security flaw, such as CVE-2021-44228, the Log4j flaw known as Log4Shell. The CVE Program assigns these IDs and keeps the public record of each disclosed flaw. CVSS, EPSS and KEV all attach to CVE IDs, so the ID is how scanners, tickets and engineers agree they mean the same problem.
Read more: CVSS vs EPSS vs KEV: How to Prioritize What Your Team Can Fix
CVSS
CVSS, the Common Vulnerability Scoring System, rates how severe a vulnerability is on a scale of 0 to 10. FIRST maintains it, and version 4.0 was released in November 2023. Scanners usually show the Base score, which describes the flaw itself, not whether attackers use it or your asset is reachable. Use it to break ties after exploitation evidence and exposure.
Read more: CVSS vs EPSS vs KEV: How to Prioritize What Your Team Can Fix
D to K
EDR
EDR, for endpoint detection and response, is software on laptops and servers that records activity, detects suspicious behavior such as files being encrypted in bulk, and lets responders isolate a device or stop a process. Many MDR providers watch EDR alerts for their customers. Decide in advance who may approve isolating a device, because that action also stops someone's work.
Read more: Security Alert Triage for Small Teams: From Queue to Resolution and Ransomware Response Plan: Who Does What, in What Order
EPSS
EPSS, the Exploit Prediction Scoring System, is FIRST's daily estimate, from 0 to 1, of the probability that a CVE sees exploitation activity in the next 30 days, plus a percentile rank. Scores move as threat data arrives, so date any value you copy into a ticket. FIRST sets no universal cut-off; pick one from your team's capacity.
Read more: CVSS vs EPSS vs KEV: How to Prioritize What Your Team Can Fix
Fractional CISO (vCISO)
A fractional CISO, also called a virtual CISO or vCISO, is a senior security leader who works for you part time, often through a firm. They bring strategy, board reporting and audit leadership when you need senior judgment but not a full-time executive. They set direction rather than chase tickets, so daily follow-up still needs an owner.
Read more: Security Team of One: What to Own, Drop and Hand Off
Human approval (read-first access)
Human approval means a person approves an AI agent's action before it happens: a ticket created, a message sent, a setting changed. Read-first access is the matching rollout pattern: the agent starts with read-only access and gains write actions one at a time, each behind approval. Together they let a small team try an agent while every change still passes through a person.
Read more: 10 Questions to Ask AI Security Vendors Before Buying
Incident response plan
An incident response plan is the written guide for a security incident: who holds command, often the security lead; who acts, usually IT and engineering; who decides on notifications, usually legal and leadership; and how you recover. NIST SP 800-61 Rev. 3, published in April 2025, replaced the 2012 Rev. 2 and ties incident response to the CSF 2.0 functions. Rehearse it before you need it.
Read more: Ransomware Response Plan: Who Does What, in What Order and Security Alert Triage for Small Teams: From Queue to Resolution
Infrastructure as code
Infrastructure as code means defining cloud resources, such as networks, storage and permissions, in version-controlled files that a pipeline applies, instead of changing settings by hand in a console. It changes how you remediate: if a misconfiguration came from code, fix the code, or the next deployment puts the bad setting back. Scanning those files before deployment catches some problems earlier.
Read more: Cloud Misconfiguration Remediation: Fix It at the Source
ISO/IEC 42001
ISO/IEC 42001 is the international standard for an AI management system, first published in 2023. It sets requirements for AI policies, roles, risk assessment and continual improvement, and organizations can be certified against it. When a vendor claims alignment, ask whether that means certification by an accredited body or a self-assessment, and what the scope covers.
Read more: 10 Questions to Ask AI Security Vendors Before Buying
KPI and KRI
A KPI, or key performance indicator, measures how well the security program performs, such as median days to fix critical findings. A KRI, or key risk indicator, measures how close the business is to its risk appetite, such as critical findings open on internet-facing systems, against a threshold leadership approves. Lead the board page with KRIs.
Read more: CISO Board Report Template: Sections, Metrics and Examples
L to R
Lean security team
A lean security team is a security function of often one to three people, at a company that has outgrown ad hoc security but has no full department. It keeps the decisions only it can make, such as risk ranking, policy and the auditor relationship, and hands execution to clear owners in IT, engineering and outside providers.
Read more: Security Team of One: What to Own, Drop and Hand Off
MSSP and MDR
An MSSP, or managed security service provider, runs security tools and monitoring on your behalf. An MDR provider, for managed detection and response, watches and investigates alerts on your endpoints and cloud, usually including hours your team is offline, and can often take first containment steps such as isolating a device. Your team still owns the fixes. Both fill monitoring gaps; neither owns your roadmap, audits or remediation follow-up.
Read more: Security Team of One: What to Own, Drop and Hand Off and Security Alert Triage for Small Teams: From Queue to Resolution
NIST AI RMF
The NIST AI RMF is the AI Risk Management Framework from the US National Institute of Standards and Technology, released as version 1.0 in January 2023 for voluntary use. It organizes AI risk work into four functions: Govern, Map, Measure and Manage. It gives a small team a shared vocabulary for vendor reviews and internal AI policy.
Read more: 10 Questions to Ask AI Security Vendors Before Buying
OWASP Top 10 for Agentic Applications
The OWASP Top 10 for Agentic Applications is a peer-reviewed list of the most critical security risks specific to AI agents, published by the OWASP GenAI Security Project in December 2025. It gives buyers and vendors a shared set of risks to test. OWASP has other Top 10 lists, so ask which one a vendor used.
Read more: 10 Questions to Ask AI Security Vendors Before Buying
Prompt injection
Prompt injection is text hidden in content an AI model reads, such as a ticket, email or web page, that tries to give it new instructions, like leaking data or taking an action nobody asked for. No vendor can promise immunity. The practical defense is limiting what an agent can do: scoped permissions, approval before writes and no ability to delete.
Read more: 10 Questions to Ask AI Security Vendors Before Buying
RACI
A RACI is a chart of who does what for each task: Responsible (does the work), Accountable (owns the outcome, one per row), Consulted (asked before) and Informed (told after). For a small team it settles who fixes, who verifies and who approves exceptions, so no finding waits for someone to claim it.
Read more: How to Verify Vulnerability Remediation After the Ticket Closes and Security Team of One: What to Own, Drop and Hand Off
Remediation SLA
A remediation SLA is the deadline you agree with engineering for fixing findings in each priority tier, shortest for the riskiest. SLA means service level agreement; customer contracts use the same term for promises such as uptime. Start near what your team already meets, tighten over time and count an SLA as met only when a rescan confirms the fix.
Read more: CVSS vs EPSS vs KEV: How to Prioritize What Your Team Can Fix and How to Verify Vulnerability Remediation After the Ticket Closes
Remediation verification (verified closure)
Remediation verification is the step that confirms a fix removed the vulnerability, usually with a rescan by the same tool, before the finding counts as closed. A ticket marked Done is a claim; a clean rescan is proof. Verified closure is the resulting status, and it is what SLA reports, board metrics and SOC 2 evidence should count.
Read more: How to Verify Vulnerability Remediation After the Ticket Closes and Vulnerability Management for a Small Security Team (2026 Guide)
Risk acceptance (exception)
Risk acceptance is a recorded decision to leave a known risk in place instead of fixing it by the agreed date. A good exception names the finding, the reason, any compensating control, who approved it and when it ends. Auditors check that exceptions were approved the way your policy says, and a recorded decision is much easier to defend than a date that slipped without one.
Read more: Vulnerability Management for a Small Security Team (2026 Guide) and Cloud Misconfiguration Remediation: Fix It at the Source
Risk appetite
Risk appetite is how much risk the business has agreed to live with, set by leadership or the board rather than the security team. In practice it becomes thresholds, such as no critical finding on an internet-facing system open past its SLA. Once it is written down, a small team can escalate against an agreed line instead of arguing case by case.
Read more: CISO Board Report Template: Sections, Metrics and Examples
RTO and RPO
The recovery time objective, or RTO, is how long a system can stay down before the outage harms the business. The recovery point objective, or RPO, is the point in time your data must be restored to, which sets how much recent data you can afford to lose. NIST SP 800-34 Rev. 1 defines both. The business sets them; restore tests show whether you meet them.
Read more: Ransomware Response Plan: Who Does What, in What Order
S to Z
Security program management (security PMO)
Security program management is running security initiatives, controls and remediation as one tracked program, with an owner, a deadline and evidence for each item. Large companies staff a security PMO, or program management office. A lean team gets most of the value from one live map of risks, owners and dates, reviewed weekly.
Read more: Security Program Management Without a PMO: A Lean Guide
Security questionnaire
A security questionnaire is a set of questions a prospective customer sends to check how you protect their data before they sign. Some buyers use a standard form such as the SIG or the CAIQ; many send their own spreadsheet or portal. Most questions map to the same controls, so a library of approved answers, with one person keeping it current, spares a small team from starting over on each deal.
Read more: How to Answer Security Questionnaires Faster (2026 Guide)
Shared responsibility model
The shared responsibility model splits cloud security between the provider and the customer. AWS describes it as the provider owning security of the cloud, meaning the infrastructure, and the customer owning security in the cloud: their data, their identity and access rules and the settings of the services they use. Misconfigured storage, permissions and network rules therefore sit on the customer's side.
Read more: Cloud Misconfiguration Remediation: Fix It at the Source
SIEM
A SIEM, for security information and event management, collects logs and events from across your systems in one place, correlates them and raises an alert when a detection rule matches. NIST's glossary defines it as centralized logging across many log types. Its value depends on tuned rules and on someone reviewing what it raises, whether the team or an MDR provider.
Read more: Security Alert Triage for Small Teams: From Queue to Resolution
SIG
The SIG, or Standardized Information Gathering questionnaire, is a licensed third-party risk questionnaire from Shared Assessments, updated every year. Buyers scope it to the risks that matter for each vendor. It is common in financial services. If your buyers send it, a completed, current SIG saves rebuilding the same answers deal by deal.
Read more: How to Answer Security Questionnaires Faster (2026 Guide) and Vendor Security Assessment: A Process a Lean Team Can Run
SOAR
SOAR, short for security orchestration, automation and response, runs playbooks you design in advance: when this alert arrives, enrich it, check threat intelligence, open a ticket. It does the same steps the same way every time, which suits work with one right answer. Work that changes case by case, such as finding an asset's owner, fits it poorly.
Read more: AI Agents for Security Teams: What They Do and Where They Fit and Security Alert Triage for Small Teams: From Queue to Resolution
SOC 2
SOC 2 is an independent attestation report on a service organization's controls, issued by a CPA firm using criteria set by the AICPA. It always covers security and can add availability, processing integrity, confidentiality and privacy. Auditors test whether you did what your own policies say, so write policies your team can follow week to week.
Read more: Is Vanta Enough for SOC 2? What It Automates and What's Left and SOC 2 Evidence Collection: What a Lean Team Still Does by Hand
SOC 2 evidence
SOC 2 evidence is the records that prove your controls exist and keep working: system settings, logs, tickets, approvals, policies and sign-offs. A compliance automation platform pulls much of the configuration evidence. Evidence of human decisions, such as access review sign-offs, change approvals and proof a flagged issue was fixed, still needs an owner to gather it.
Read more: SOC 2 Evidence Collection: What a Lean Team Still Does by Hand
SOC 2 Type 1 vs Type 2
A SOC 2 Type 1 report checks that controls are suitably designed at one point in time. A Type 2 report also tests that they operated effectively over a period, often several months, called the observation window, and the auditor samples evidence from across it. When reviewing a vendor, ask which type it holds.
Read more: 10 Questions to Ask AI Security Vendors Before Buying and SOC 2 Evidence Collection: What a Lean Team Still Does by Hand
SSVC
SSVC, Stakeholder-Specific Vulnerability Categorization, is a decision-tree method from CERT/CC at Carnegie Mellon University's Software Engineering Institute, working with CISA. It asks about exploitation, exposure, automatability and impact, then points to an action; CISA's version ends in Track, Track*, Attend or Act. A small team can borrow its questions without running the full tree.
Read more: CVSS vs EPSS vs KEV: How to Prioritize What Your Team Can Fix
Tabletop exercise
A tabletop exercise is a discussion-based rehearsal where the people named in your incident response plan walk through a realistic scenario, such as ransomware, and say what they would do at each step. Nothing on live systems changes. CISA publishes exercise packages with scenarios and discussion questions. The useful output is the list of gaps found, each with an owner and a date.
Read more: Ransomware Response Plan: Who Does What, in What Order
Third-party risk management (TPRM)
Third-party risk management, or TPRM, is the ongoing work of finding, assessing and tracking the risk that vendors and other outside parties bring, from the review before signing through reassessment and offboarding. Security gathers the evidence and finds the gaps; a named risk owner decides whether the risk is acceptable. Tiering vendors keeps a lean team's effort in proportion to each vendor's risk.
Read more: Vendor Security Assessment: A Process a Lean Team Can Run
Trust center
A trust center is a public or gated web page where buyers can read your security overview and request documents such as your SOC 2 report, often under NDA. Buyers serve themselves the standard material, so fewer questionnaires reach the security team. Keep it current: an expired report or stale policy raises more questions than it answers.
Read more: How to Answer Security Questionnaires Faster (2026 Guide)
Trust Services Criteria (CC6 logical access)
The Trust Services Criteria are the AICPA's criteria that a SOC 2 report measures, dating from 2017 with points of focus revised in 2022. The CC6 series covers access. CC6.1 is logical access security for systems and data; CC6.2 is registering, authorizing and removing users; CC6.3 is role-based access, least privilege and segregation of duties. Access reviews are the usual evidence.
Read more: Quarterly Access Reviews for SOC 2: A Step-by-Step Guide
User access review
A user access review is a periodic check where the right reviewer confirms that each person's access to a system is still needed, and access that is not needed is removed. SOC 2 does not fix the frequency; auditors test what your own policy says, often quarterly. The review counts only when flagged access is removed and you can prove it.
Read more: Quarterly Access Reviews for SOC 2: A Step-by-Step Guide
Vendor tiering
Vendor tiering is sorting vendors into risk levels, usually by the data they can reach, the access they hold and how much the business depends on them, so the depth of each review matches the risk. A vendor holding customer data may need a SOC 2 report and a full questionnaire; a low-risk tool may need a short check. Each tier should set when the vendor is reassessed.
Read more: Vendor Security Assessment: A Process a Lean Team Can Run