NINE OUT OF TEN COMPANIES HAVE BEEN BURNED BY THEIR OWN AI AGENTS AND MOST STILL CAN’T STOP IT FROM HAPPENING AGAIN A new VentureBeat survey of 145 enterprise respondents has turned up a number that should alarm anyone overseeing AI deployment in a large organization: 88 percent of enterprises reported at least one AI agent security incident in the past year. The number reflects what insiders have been saying for months in closed rooms, which is that autonomous AI agents deployed in production are creating security exposure that companies have not yet figured out how to manage. The most common threat is not a sophisticated external attack. The most cited control failure, flagged by 49 percent of respondents, is shadow AI: unauthorized agentic pipelines built on corporate credit cards, running outside any central oversight, pulling company data into models that have not been reviewed or approved. Another 25 percent had been hit by runaway agents spinning in infinite loops and racking up unexpected charges. The governance picture is equally grim. Only 38 percent of companies have a central team governing AI. A fifth of respondents said each platform team governs its own AI independently. One in six reported that no specific role holds formal accountability at all. These are not fringe companies experimenting carelessly. These are 100-plus employee organizations running production AI in 2026. The agents are in the building. The oversight is not. Keywords: AI agent security incidents, enterprise AI risk, shadow AI, AI governance failures