CYBER Discuss an assessment

Research

AI security findings.

Examples of findings from AI security assessments, showing what we tested, what happened and why it mattered.

Shared agent workspace exposed files across tenants

Missing organisation ID exposed another organisation's document

Resuming another tenant's conversation exposed its history

Traveller account could create shared instructions followed by the assistant

Fabricated documents influenced board advice through RAG

Another tenant could read an agent result and submit feedback