A live operational risk: Why AI agents are outrunning your security (www.techradar.com)

A recent Deloitte report highlights a significant governance gap in the use of autonomous AI agents, with only 21% of organizations reporting mature governance frameworks despite 73% expressing concerns over AI security and data privacy. While many companies are rapidly deploying AI agents into live operations—23% are using them moderately as of now, with expectations that this will rise to nearly 75% in two years—effective governance has not kept pace. This discrepancy poses an operational risk, as current governance frameworks, designed for human operators and deterministic software, are ill-equipped for the complex, continuous decision-making behavior of AI agents. The report indicates that organizations often engage in "governance theater," where policies exist on paper but lack the necessary technical infrastructure for enforcement in real-time. To mitigate risks, experts recommend comprehensive auditing of agent activities, precise mapping of permission scopes, and treating AI systems as first-class identities with defined ownership and monitoring. Without these crucial components in place, companies face the danger of unchecked AI operations and the compounded negative consequences of inadequate governance as deployments scale. Building robust, enforceable governance is essential for the continued safe integration of AI into business processes.