Cyber Risks · The Hacker News

AI Agent Lifecycle Risks

Summary: Security leaders largely lack visibility into how many AI agents run inside their organizations, with 109 non-human identities now existing for every human and 82% of organizations having discovered shadow AI no one approved. Agents are created in minutes, accumulate permissions as they operate, and retain access after they are no longer needed, while only 21% of companies have a proper way to shut one down.

Bottom line: With 109 non-human identities per human and only 21% of companies able to shut an agent down, the access your agents quietly accumulate outlives the work they were built to do.

Read the original at The Hacker News ↗