Rubrik Warns of Growing Security Gaps as AI Agent Adoption Accelerates
A Rubrik Zero Labs report indicates that AI agent adoption is outpacing security measures. 86% of IT leaders expect AI agents to overwhelm security guardrails within a year, while visibility gaps and the rise of a 'shadow workforce' create significant risks and recovery concerns.
📋 Article Processing Timeline
- 📰 Published: May 19, 2026 at 20:10
- 🔍 Collected: May 19, 2026 at 11:31
- 🤖 AI Analyzed: May 28, 2026 at 17:24 (221h 52m after Collected)
Rubrik, a leader in security and AI operations, today released the latest research report from Rubrik Zero Labs. The study reveals that AI agent adoption in enterprises is moving faster than the development of security measures. Organizations are operating agent-based systems without sufficient governance controls, leading to a gap between innovation and security.
The survey of 1,600 IT and security leaders revealed key findings: 86% of respondents (83% in Japan) predict that the surge in AI agents will surpass their security guardrails within the next year. Only 23% (38% in Japan) reported being able to fully monitor active AI agents in their environment. As a result, identities linked to critical data and decision-making are not being adequately protected.
This challenge is worsened by the explosion of identities. Non-human identities associated with agents are increasing faster than corporate tracking can keep up, a phenomenon researchers call the 'shadow workforce.' These identities often operate with persistent access and limited oversight, creating new pathways for misuse, breaches, and lateral movement.
Expectations for AI agent operations are also wavering. Over 80% (86% in Japan) stated that the manual management burden outweighs the efficiency gained from agent adoption. Furthermore, 88% (84% in Japan) reported lacking the capability to undo specific actions taken by AI agents without resetting the entire system. Approximately 90% expressed concern about meeting recovery goals.
The threat landscape is accelerating, with half of respondents expecting agent-based systems to account for the majority of attacks within a year. AI agents are shortening attack times and increasing scale, blurring the lines between internal risks and external breaches.
Kavitha Mariappan, Chief Transformation Officer at Rubrik, stated: 'AI adoption is outpacing the ability to control it. As decision-making shifts from humans to machines, maintaining operational safety in autonomous environments is a critical challenge.' AI strategy is now inseparable from resilience strategy, making identity verification and visibility prerequisites for safe agent system adoption.
The survey of 1,600 IT and security leaders revealed key findings: 86% of respondents (83% in Japan) predict that the surge in AI agents will surpass their security guardrails within the next year. Only 23% (38% in Japan) reported being able to fully monitor active AI agents in their environment. As a result, identities linked to critical data and decision-making are not being adequately protected.
This challenge is worsened by the explosion of identities. Non-human identities associated with agents are increasing faster than corporate tracking can keep up, a phenomenon researchers call the 'shadow workforce.' These identities often operate with persistent access and limited oversight, creating new pathways for misuse, breaches, and lateral movement.
Expectations for AI agent operations are also wavering. Over 80% (86% in Japan) stated that the manual management burden outweighs the efficiency gained from agent adoption. Furthermore, 88% (84% in Japan) reported lacking the capability to undo specific actions taken by AI agents without resetting the entire system. Approximately 90% expressed concern about meeting recovery goals.
The threat landscape is accelerating, with half of respondents expecting agent-based systems to account for the majority of attacks within a year. AI agents are shortening attack times and increasing scale, blurring the lines between internal risks and external breaches.
Kavitha Mariappan, Chief Transformation Officer at Rubrik, stated: 'AI adoption is outpacing the ability to control it. As decision-making shifts from humans to machines, maintaining operational safety in autonomous environments is a critical challenge.' AI strategy is now inseparable from resilience strategy, making identity verification and visibility prerequisites for safe agent system adoption.
FAQ
AIエージェントの監視における主な課題は何ですか？
IT環境内でアクティブなAIエージェントを完全に監視できていると回答したのは、グローバルでわずか23%（日本は38%）に留まっており、深刻な可視化不足が指摘されています。
「シャドーワークフォース」とはどのような概念ですか？
AIエージェントに紐づく、企業による追跡やガバナンスが追いつかない速度で急増している「非人間アイデンティティ」を指し、新たな攻撃経路となるリスクを孕んでいます。
AIエージェントの導入は業務の効率化に寄与していますか？
回答者の80%以上（日本は86%）が、AIエージェント導入による効率化よりも、人手による管理負担の方が大きいと回答しています。
AIエージェントの誤動作に対する復旧能力はどうなっていますか？
88%（日本は84%）の組織が、システム全体をリセットせずにAIエージェントの特定のアクションを元に戻す機能を備えていないことが判明しました。
今後のサイバー攻撃はどのように変化すると予想されていますか？
回答者の約半数が、今後1年以内にはエージェント型システムを用いた攻撃が大半を占め、攻撃の高速化と規模の拡大が進むと予想しています。