Jane Edwards 
- Trellix has published new research on AI-driven insider risk
- Most organizations lack the visibility and policies to manage AI agents
- The 2026 Homeland Security Summit will feature several CISA leaders as speakers
A Trellix report has found that 95 percent of surveyed security leaders encountered an insider incident tied to an artificial intelligence tool or autonomous system over the past 12 months.
As Trellix research points to AI as a growing source of insider risk, Cybersecurity and Infrastructure Security Agency leaders will share their perspectives on protecting the nation’s critical infrastructure at the Potomac Officers Club’s 2026 Homeland Security Summit on Nov. 10. Bobby Thompson, acting associate director for threat hunting at CISA, will deliver a keynote address, while CISA’s Lauren Wind and Daryle Hernandez will join panel discussions on data and identity integration and trust in homeland security technologies. Register now to join the conversation.
The cybersecurity company said Tuesday that its Mind of the CISO: The New Insider report draws on a survey of 600 chief information security officers and senior security leaders.
How Does AI Insider Risk Affect Organizations?
According to the report, AI-driven actions account for nearly one-third, or 28 percent, of insider-related incidents, with each incident costing organizations an average of $2.4 million. Respondents cited data loss as the most common consequence at 56 percent, followed by operational disruption at 49 percent, compliance or legal violations at 48 percent and damage to reputation and customer confidence at 46 percent.
The survey also found that 78 percent of organizations now count AI agents, autonomous systems and machine identities within the scope of insider risk, and 42 percent named this area their top security priority for the next one to two years. However, 52 percent have yet to fully put an insider risk strategy in place, while 94 percent expressed concern about AI actions within their security environments.
What Are the Report’s Other Key Findings?
- Sixty-three percent of CISOs reported insufficient visibility into employee AI tool use, and 99 percent see a need to improve visibility across users, devices and environments.
- Organizations need an average of 15 hours to detect and respond to insider-related incidents.
- Eighty-three percent of respondents believe AI is helping attackers find and exploit vulnerabilities faster, while 78 percent said traditional security controls are losing effectiveness against such attacks.
- Seventy-nine percent anticipate an AI-assisted attack on trusted identities or systems within the next year.
- Ninety-nine percent view AI and automation as essential to insider risk management over the next 12 to 24 months, and 41 percent plan to prioritize spending on AI-based detection and response.
- Ninety-one percent of CISOs said their accountability for insider risk has increased over the past year, and 69 percent are concerned about personal liability following an incident.
- Seventy-four percent reported that senior leadership and boards lack a clear understanding of insider risk.
- Sixty-two percent said their organizations are not fully prepared for AI-related risks, and 68 percent said existing policies fall short, with most respondents seeking clearer regulatory frameworks, compliance requirements and best practices.
The latest findings build on Trellix’s Mind of the CISO: The Future of Cyber Resilience report, released in December 2025, in which 97 percent of over 500 surveyed CISOs said hybrid infrastructure offers greater resilience than cloud-only or on-premises models. Respondents in that survey identified ransomware and extortion, autonomous AI-enabled attacks and threats to operational technology environments as the main factors reshaping their security strategies.
What Did Trellix & CISA Leaders Say About AI Insider Risk?
Trellix CISO David Soto said organizations should govern AI agents as privileged identities while building agentic security operations centers capable of responding at machine speed.
“A 15-hour detection-to-response window isn’t just a gap; it’s an open invitation,” Soto noted. “When AI agents operate inside your trusted boundary at machine speed, the resilient organizations will be the ones governing them like the privileged identities they are, while simultaneously building an agentic SOC to respond at the speed of the threat, not a human workflow.”
Clayton Romans, associate director for the Joint Cyber Defense Collaborative at CISA, said the agency is working with government and private sector partners to develop guidance that helps organizations manage AI-related risks.
“Autonomous agents now operate inside enterprise trust boundaries with access equivalent to privileged human users, and in many cases, without equivalent governance,” Romans said. “The risk is already here; the time for leaders and cyber defenders to act is now.”
