AI interface & API keys
The AI interface and customer-supplied API keys are protected against prompt injection, attempts to bypass pseudonymisation rules, searches for re-identifying data and misuse of document generation functions.
The AI interface and customer-supplied API keys are protected against prompt injection, attempts to bypass pseudonymisation rules, searches for re-identifying data and misuse of document generation functions.
The interface incorporates controls, isolation rules and protection against prompt injection and abnormal behaviour. They prevent users or malicious instructions from diverting tools to access information that should not be exposed.
In Layer, the models, agents and integrations you choose access only pseudonymised documents you have explicitly marked as processed and approved. They work within this authorised document scope. The software resolves entities locally.
When AI searches the web, allow and block lists control permitted sites. You define the web search scope: only approved domains may be consulted, while unwanted sources remain excluded.