campustechnology.com·18 november 2026
Survey: Half of Companies Confirm Unauthorized AI Agent Actions -- Campus Technology

AI Actions
AI agents
AI configuration
AI intervention
AI inventory
AI monitoring
AI permissions
AI security
AI transactions
AI visibility
API security
automated controls
data modification
Salt Security
unauthorized actions
campustechnology.com
A new report by Salt Security reveals that 49.8% of organizations experienced unauthorized AI agent actions in the past year. The survey of 380 security professionals found that 53.9% of incidents involved external API calls, while 41.9% reported data modification or deletion. Only 12.5% of organizations could consistently trace all agent actions, and just 17.5% could automatically stop an AI agent mid-action. Salt Security recommends maintaining inventories of AI agents, assigning limited permissions, and implementing automated monitoring to prevent high-risk actions.

