FirmOps terms, on the URLs that already rank.

Short definitions

The trust terms, in one line each.

MCP (Model Context Protocol)

The project's own docs call MCP "an open-source standard for connecting AI applications to external systems." For a firm, that means Claude or ChatGPT can reach case management, documents, and email through one connector pattern. MCP is plumbing, not a security guarantee. Scoped permissions, logging, and approvals are still implementation work. [1]

Hallucination

NIST files this risk under "confabulation": "the production of confidently stated but erroneous or false content (known colloquially as “hallucinations” or “fabrications”) by which users may be misled or deceived." In a firm, the fix is not a better prompt. It is a person checking the output before it leaves the building. [2]

Approval gate

A step where a named person approves AI output before it is sent, filed, or written back to a system of record. It is the operational answer to hallucination risk.

Read-first

Deployment order where the brain gets read access across approved systems before it is allowed to write anything.

Shadow AI

Staff using AI tools the firm has not approved, usually with client data pasted into a personal account.