Glossary | Prompt Security
# AI Security Glossary
Get familiar with key terms and concepts in the AI Security space.
Agentic AI
Agentic AI refers to AI systems built to act autonomously toward a goal, planning multi-step tasks, choosing tools, and executing actions, rather than simply responding to a single prompt.
AI Acceptable Use Policy
An AI Acceptable Use Policy (AUP) sets clear guidelines for how employees, contractors, and partners can responsibly use AI tools and agents at work, including what's approved, what data can be shared, and who's accountable when something goes wrong.
AI Gateway
An AI gateway is a centralized layer between an organization's users or applications and the AI models or services they call, letting security teams monitor, control, and secure that traffic from one place instead of instrumenting every application separately.
AI Inventory
An AI inventory is the complete, actively maintained record of every AI tool, model, and agent in use across an organization, including the ones IT never approved.
AI Pipeline
An AI pipeline is a type of data pipeline built specifically to support AI use cases, giving teams a structured, repeatable way to build and run AI systems from data collection through deployment and monitoring.
AI Red Teaming
AI red teaming tests an AI application or agent against real adversarial techniques, prompt injection, jailbreaks, tool misuse, before an actual attacker does.
AI Security
AI security covers the measures, policies, and controls that protect an organization from the risks introduced by its use of AI, spanning employee tool usage, first-party applications, and increasingly, autonomous agents.
Artificial General Intelligence (AGI)
AGI refers to a hypothetical AI system capable of performing any intellectual task a human can, without being specifically trained for it. It remains theoretical, with no consensus definition or benchmark.
Chief AI Officer (CAIO)
A Chief AI Officer (CAIO) is a senior executive responsible for an organization's AI strategy, balancing the pressure to adopt AI quickly against the risks that adoption introduces.
Data Loss Prevention for AI (DLP for AI)
DLP for AI extends data loss prevention beyond files and email to the places sensitive data now moves through AI systems: prompts, model outputs, and agent-to-tool traffic that conventional DLP tools were never built to inspect.
Data Privacy Risks
Data privacy risk in the AI era covers every way sensitive information can end up somewhere it shouldn't, from an employee pasting confidential data into a chatbot to an autonomous agent passing data between connected tools.
Deepfake
Deepfakes are AI-generated or AI-altered visual or audio representations of identifiable people, most often created to make it appear as though someone said or did something they didn't.