The world of AI agent security is a complex and evolving landscape, and this article delves into the latest findings from a VentureBeat Pulse Research survey, shedding light on the challenges and opportunities in securing AI agents. The survey, conducted in July 2026, gathered insights from 116 qualified enterprise respondents, offering a comprehensive view of the current state of AI agent security practices and perceptions.
The Growing Threat of AI-Armed Attackers
The survey reveals a concerning trend: as many enterprises believe AI-armed attackers are ahead of their defenses (30%) as those who believe their defenses are ahead (30%). This even split highlights the growing concern among organizations about the potential threat posed by AI-enabled adversaries. Interestingly, among those who have experienced confirmed incidents or near-misses, a staggering 39% believe attackers are ahead, compared to 20% among those who have not been affected. This stark contrast underscores the impact of real-world experiences on security perceptions.
The Containment Gap
One of the critical findings is the 'containment gap' in AI agent security. Despite the majority of enterprises (53%) having agentic AI systems in production, only 18% isolate their highest-risk agents, and a mere 8% pair enforcement with isolation. This gap means that when security controls fail, the potential blast radius remains wide, as credential sharing persists across 63% of fleets. The reliance on borrowed, provider-native controls, as evidenced by the primary security layer choices (92% using hyperscaler or model provider-native controls), further exacerbates the issue.
Satisfaction and Churn Intent
Surprisingly, despite the security concerns, satisfaction with current agent security tooling is at a series high of 4.29 out of 5. However, this high satisfaction is accompanied by a significant churn intent. Three-quarters of enterprises plan to adopt, add, or replace agent security tooling within 12 months, with 30% intending to make changes within the next quarter. This indicates that while organizations are content with their current solutions, they are also actively seeking improvements to address the identified security gaps.
The Role of Budgets
Budget allocation for agent security is gradually increasing, but it remains a modest slice of the overall security budget. The most common allocation is 6-10% (44%), and a meaningful minority (35%) now devote more than a tenth of their budget to agent security. This shift in spending suggests that enterprises are recognizing the importance of investing in security measures to mitigate the risks associated with AI agents.
The Need for Isolation and Governed Identity
The survey highlights the critical need for isolation and governed identity in AI agent security. While enterprises are building agent security in earnest, they are not prioritizing containment. The lack of isolation and governed identity controls is evident in the consideration set, where only 10% of enterprises include agent-identity products in their plans. This oversight is particularly concerning, given the direct implications of incidents on the need for these controls.
The Way Forward
The survey's findings emphasize the importance of addressing the containment gap in AI agent security. Enterprises must move beyond relying solely on provider-native controls and dedicated security vendors. By building isolation and governed identity controls deliberately, organizations can create a more robust defense-in-depth architecture. The open question for future waves is whether enterprises will proactively address these gaps or whether confirmed incidents will force them to take action.
In conclusion, the world of AI agent security is at a critical juncture, with the need for containment and governed identity becoming increasingly apparent. Enterprises must take a proactive approach to security, recognizing that the current state of affairs may not be sustainable in the face of evolving threats. The future of AI agent security depends on the collective efforts of organizations to build a more secure and resilient infrastructure.
