AI compliance and observability glossary

Shadow AI is the use of consumer AI tools — ChatGPT, Claude, Gemini, Copilot — by employees without IT approva…

AI observability is the practice of capturing every LLM call, scoring it for quality, and retaining the record…

LLM observability is the model-layer subset of AI observability. It covers prompt-response capture, token and …

AI red teaming is the practice of probing AI systems — particularly LLMs — for jailbreaks, prompt injection, p…

An AI guardrail is a real-time policy layer that intercepts LLM inputs and outputs to detect, block, or redact…

Model drift is the silent decline of an AI model's quality, accuracy, or behavior over time. In LLMs and agent…

Prompt injection is a class of attack where adversarial input is crafted to redirect a large language model's …

Tool calling, sometimes called function calling, is the capability that lets a large language model invoke ext…

PII redaction is the process of detecting personal identifying information in text and removing, masking, or r…

HIPAA Safe Harbor is one of two methods the HIPAA Privacy Rule provides for de-identifying protected health in…

LLM-as-Judge is an evaluation technique where an independent large language model scores another LLM's output …

Defined on the Traces page.

Step-by-step record of an agent's decision.

PII scrubbed before storage.

What it is and how Prism blocks it.

Pre-deployment risk assessment.

Voluntary US framework, mapped to Prism.

Annex III scope, with Prism mapping.

FRB / OCC model risk management.

AI management system standard.

HIPAA's 18 identifier categories.

What Prism X does at the browser.