Hallucination
The generation of outputs by an AI model that are plausible-sounding but factually incorrect, fabricated or unsupported by input data or training sources.
In plain language
When an AI confidently makes something up. It might state a false statistic, cite a research paper that doesn't exist or invent a historical event, all while sounding completely certain.
Why this matters
Hallucination is one of the most significant risks in deploying generative AI and demands explicit governance controls. Your organisation must detect, mitigate and disclose hallucination risks wherever AI outputs inform business decisions, customer interactions or regulatory submissions.
Relevance
GovernanceHallucination risk must be explicitly managed in governance frameworks to protect organisations from reputational, legal and regulatory harm.
Related terms
Putting hallucination into practice in your organisation?
Ready to transform your AI strategy?
Partner with Australia's AI strategy and governance specialists. From adoption roadmaps to ISO 42001 audit readiness.