AI agent adoption is accelerating across organisations, but Splunk’s State of Observability report highlights a widening gap between rollout and accompanying security controls. The article notes that many businesses deploy AI agents rapidly without clear ownership or accountability for monitoring, with 38% of IT/Ops respondents lacking defined responsibility and 42% saying a different team is responsible for observing AI.
This fragmentation across software engineering, infrastructure management, web development and information security hampers effective oversight, risking hallucinations, data exposure and other AI-related issues.
Evidence from the survey shows a tension between desire and capability: 85% of respondents value visibility into AI hallucinations, yet only 58% can effectively monitor for them. Organisations may struggle to recognise when an agent errs or acts inappropriately, potentially enabling a malicious actor to exploit the system. Cisco’s Jeetu Patel emphasised the need to converge security and observability data and to rely on common data platforms for real-time detection and response.
The report does acknowledge potential benefits when AI is deployed well: 86% reported faster detection of IT or security issues, with 36% describing the improvement as significant. Nevertheless, 14% said AI deployments made issues harder to detect or resolve, underscoring the requirement for robust governance, cross-team collaboration and scalable monitoring to realise reliable, safe agent use.