Shadow AI refers to AI tools that employees use for work tasks without the approval or knowledge of their organization. Examples are private accounts with chat services, browser extensions with AI features and AI functions in existing software that someone switches on without authorization. The term is modelled on shadow IT.
The risk lies in what leaves the organization: confidential documents, personal data or source code end up with providers that have no data processing agreement, and nobody knows which systems are in use. An AI Usage Policy with approved tools and clear rules for inputs has proven effective, supported by an AI Register and training. Approved alternatives matter, because experience shows that a plain ban pushes use onto private devices.