Glossary
AI governance and AI security terms, defined plainly
The terms used across this site and in most AI governance conversations, each in one or two sentences, with a link to the page that goes further.
- Admission gateway
- A control point that decides whether an AI request may proceed before any model provider is contacted. A refused request is never sent, so it reaches no provider and incurs no provider cost. Read more
- AI agent
- Software that uses a model to decide on actions and carry them out, usually by calling tools. Its risk depends on what those tools can do: read, write, delete, pay or send data outside the organisation. Read more
- AI FinOps
- The practice of making AI spend visible, attributable and accountable: measuring tokens and cost by team and model, setting budgets and choosing models deliberately. Read more
- AI gateway
- A proxy between AI callers and model providers that routes, logs or controls requests. Some gateways focus on routing and caching for developers; an admission gateway focuses on deciding whether a request may proceed. Read more
- AI governance
- The decisions, records and controls through which an organisation knows which AI it uses, decides what is acceptable, enforces that decision and can prove it. Read more
- AI literacy
- Under the EU AI Act, the skills, knowledge and understanding that let providers, deployers and affected persons make an informed deployment of AI systems and be aware of their opportunities and risks. Read more
- AI security posture management (AI-SPM)
- Tools that assess the configuration and exposure of AI workloads and models, usually in cloud environments. It is a posture view rather than a control at the point of each request. Read more
- AI system inventory
- A register of the AI systems an organisation uses, with owner, purpose, vendor, models, data and status for each. Also called an AI register. Read more
- Audit package
- A downloadable file containing every governance record and evidence entry about one AI system or vendor, as stored, with a statement of what it does not prove. Read more
- Data residency
- A requirement that data is processed or stored only in declared regions. Applied to AI, it decides which provider endpoints a request may reach. Read more
- Deployer
- Under the EU AI Act, a natural or legal person that uses an AI system under its authority, except for personal non-professional activity. Most enterprises are deployers of the AI they buy. Read more
- Evidence log
- An append-only record in which each entry carries the hash of the one before, so that any later edit is detectable. Read more
- High-risk AI system
- Under the EU AI Act, an AI system classified as high-risk under Article 6, either as a safety component or product covered by the legislation in Annex I or as a use listed in Annex III. These systems carry the most extensive obligations in the Act. Read more
- Live dispatch
- The switch that lets admitted requests actually call the model provider. With it off, a policy can be tested at no provider cost. Read more
- MCP server
- A Model Context Protocol server. It exposes tools and data from enterprise systems to AI models and agents. Read more
- Model Context Protocol (MCP)
- An open protocol that standardises how AI applications connect to external tools and data sources. Read more
- Pinned basis
- The exact classification, reviews, vendor approval and inventory snapshot on which an approval was requested, kept so that any later change is visible. Read more
- Prompt DLP
- Data loss prevention applied to the content of an AI request: the prompt, system messages, tool arguments and tool results. Read more
- Prompt injection
- An attack in which instructions hidden in input or retrieved content cause a model to ignore its intended instructions or take unintended actions. Read more
- Provider
- Under the EU AI Act, the party that develops an AI system or model, or has it developed, and places it on the market or puts it into service under its own name. Read more
- Sanctioned provider list
- The written list of AI providers and services an organisation has approved. Anything outside it is, by definition, shadow AI. Read more
- Self-hosted
- Software that runs on infrastructure the customer controls and operates, with no vendor cloud in the path. Read more
- Separation of duties
- The rule that the person who requests an approval cannot be the person who grants it. Read more
- Shadow AI
- AI services, features or agents used inside an organisation without having been approved through its governance process. Read more
- Single-tenant
- A deployment that serves one organisation only, with no other customer data beside it. Read more
See MERIDVAR refuse a request on your own network
A working session with the founder, on a host you control.
What happens next
- You send the formYour request goes straight to the founder.
- We reply by emailWe agree a time and what you want to see.
- We run a working sessionWe install the appliance with you, route a test request through it and read the evidence it leaves.