The State of AI Agent Security: 73% of CISOs fear AI agent risks but only 30% are ready

NeuralTrust publishes a global survey revealing how security leaders are responding to the rise of autonomous AI systems and the new discipline of Agentic Security.

NEW YORK, Nov. 6, 2025 /PRNewswire/ -- As enterprises accelerate AI adoption, a new global survey by NeuralTrust reveals a growing gap between AI innovation and security readiness.

According to The State of AI Agent Security 2026, 73% of CISOs are very or critically concerned about AI agent risks, but only 30% have mature safeguards in place to protect them.

The study, based on responses from more than 160 CISOs and security leaders worldwide, finds that most organizations are deploying AI agents faster than they can secure them. NeuralTrust's maturity model places 46% of enterprises in the Reactive tier and 29% in Managed, with fewer than 10% achieving Proactive governance.

"AI agents are now part of enterprise operations, but the security controls protecting them are still catching up," said Joan Vendrell, Co-Founder and CEO of NeuralTrust. "Our findings show that Agentic Security has become one of the most urgent and complex challenges in modern cybersecurity."

AI agent incidents are already happening

The report found that 1 in 5 organizations has experienced at least one AI agent–related security breach, mainly driven by prompt injection and data exposure.

Among affected enterprises, 40% estimate financial losses between $1 million and $10 million, and 13% expect impacts exceeding $10 million, levels comparable to large-scale ransomware incidents.

Key findings

Most enterprises are still relying on traditional security controls not built for autonomous systems. As a result, visibility is improving, but control remains limited.

Here's what the survey found:

• 42% use activity monitoring to observe AI behavior,
• 38% rely on access control to limit permissions,
• 31% apply data loss prevention tools to protect sensitive data,
• Only 19% conduct adversarial testing to simulate real attacks,
• Just 16% validate their AI supply chain to detect inherited risks.

Alarmingly, one in four organizations (25%) has no AI-specific controls at all, highlighting how unprepared many teams remain for the shift from human-triggered systems to autonomous AI agents.

AI Agent Security will define enterprise resilience

By 2028, one in three enterprises will operate more than 500 AI agents, and by 2030, over half will. At the same time, 80% of organizations will fall under AI-specific regulation, and three-quarters will employ dedicated AI security specialists.

As this shift accelerates, Agentic Security will become a strategic foundation of enterprise trust.

About the report

The State of AI Agent Security 2026 is a global survey conducted by NeuralTrust between August and September 2025, capturing responses from CISOs, Chief Data Officers, and security executives across North America, Europe, Asia-Pacific, and Latin America. The study examines how enterprises are securing, monitoring, and governing AI agents as adoption scales and regulation tightens.

The next wave of AI adoption won't be defined by speed, but by trust. Agentic Security is emerging as the discipline that will decide which organizations can harness autonomy safely and lead with confidence.

Download for free here.

About NeuralTrust

NeuralTrust is the leading platform for securing and scaling AI agents and LLM applications. Recognized by the European Commission as a champion in AI security, NeuralTrust helps enterprises protect critical AI systems through runtime protection, threat detection, and compliance automation.

Our mission is to make AI adoption measurable, secure, and reliable turning AI Agent Security into a strategic advantage.

Learn more at neuraltrust.ai

SOURCE NeuralTrust