A recent survey by Gravitee reveals that more than half of AI agents deployed by large firms in the US and UK lack proper monitoring and security. The study, which surveyed 750 IT executives, estimates there are over three million AI agents in use, with 53% of them ungoverned. This raises significant concerns, as 88% of respondents reported experiencing or suspecting incidents related to AI agents in the past year. Rory Blundell, CEO of Gravitee, emphasizes that while AI agents can boost productivity, the risks of unmonitored agents can turn them into liabilities. Experts like David Shipley of Beauceron Security compare the current situation to the Titanic disaster, highlighting that organizations are aware of the dangers but continue to rely on flawed assumptions about AI governance. Shipley argues that all AI agents have the potential to go rogue, and underestimating this risk is a dangerous mistake. Manish Jain from Info-Tech Research Group points out that many organizations lack visibility into their AI agents, with unaccounted agents often emerging outside of traditional IT oversight. He warns that the real issue is "invisible AI," which can operate unchecked and become a significant threat. Jain advocates for continuous oversight and a shift in perspective regarding AI agents. Many organizations treat these agents as helpful tools, but they often come with broad access and undefined accountability, potentially becoming insider threats. Establishing tiered access for AI agents is essential to mitigate risks, similar to how one wouldn’t hand out house keys indiscriminately. The findings of this study underscore the urgent need for improved governance as AI adoption accelerates.