Hallucination
A hallucination is when an AI states something confidently that isn't true — a fact, a citation, a piece of data — because it generated plausible-sounding text rather than checked a real source.
LLMs are built to produce fluent, likely-sounding continuations of text, not to verify facts. When a model doesn't actually know something, it will often generate a confident-sounding answer anyway rather than say so — which is the core reason AI output needs to be grounded in something real.
Assistant
Masterforce agents work against real workspace content — actual documents, tasks, and records read directly via file-first reads — rather than free-form recall, and a Checkout lock keeps two agents from acting on stale or conflicting versions of the same thing. Grounding in real, current data is the main defense against an agent inventing an answer.