Glossary
Updated
While traditional IT focuses on protecting data inputs, AI governance must also manage outputs. Output risk includes hallucinations (plausible-sounding lies), algorithmic bias that leads to discriminatory decisions, and the inadvertent creation of content that infringes on third-party copyrights or trademarks.
Example
A marketing team uses an unsanctioned AI to generate social media copy. The AI hallucinates a non-existent 50% discount, leading to a surge of angry customers and a legal requirement to honor the price.