CSA Survey: 92% Security Leaders Worry About AI Agent Risks, 73% Already Hit by AI Attacks

Published: June 24, 2026 ｜ Source: Cloud Security Alliance (CSA) + Darktrace's "The State of AI Cybersecurity 2026" report, surveying over 1,500 security leaders

A Wake-Up Call from 1,500 Security Leaders: AI Attacks Are Already Here

The newly released "The State of AI Cybersecurity 2026" report by the Cloud Security Alliance (CSA) and cybersecurity firm Darktrace collected responses from over 1,500 CISOs, IT leaders, and security practitioners. The results reveal a troubling reality: AI-driven cyberattacks are no longer a "future threat" — they are actively impacting enterprises right now.

Key Numbers Paint a Clear Crisis Picture

Several striking data points from the report勾勒出 the current state of the AI security battlefield:

• 73% of security leaders say AI-driven threats have already had a significant impact on their organization — nearly three-quarters of enterprises are facing direct AI attack consequences
• 92% of companies worry about the security implications of AI agents in the workplace — employees embedding autonomous AI agents into daily workflows introduce data exposure, unauthorized actions, and opaque decision-making risks
• 44% are "extremely" or "very concerned" about the security of third-party LLMs (like Copilot, ChatGPT) — C-level awareness is rising
• 61% cite sensitive data exposure as their top concern, 56% fear compliance violations — these risks bring the fastest and most severe fines and reputational damage

AI Is Reshaping Security Operations: 85% of Human SOCs Going the Way of the Dodo

The report reveals a surprising but clear trend: enterprises are rapidly abandoning in-house SOCs (Security Operations Centers), outsourcing SOC services to MSSPs (Managed Security Service Providers).

A staggering 85% of security leaders prefer relying on MSSPs for SOC services rather than maintaining internal teams. The reasons are pragmatic — building a SOC requires round-the-clock staffing, continuous technology updates, and enormous operational costs. While 96% of cybersecurity professionals agree AI can significantly improve their work speed and efficiency, only 35% use unsupervised machine learning, and just 14% allow AI to take independent remediation actions in the SOC without human intervention.

"Know to Change" vs. "Actually Do": 97% Adopt AI Security Tools, But Only 14% Dare to Let Go

Another staggering gap: Fortinet statistics show 97% of organizations have adopted or plan to adopt AI-powered cybersecurity solutions. The question isn't "whether to use AI," but rather "can we trust it?"

Only 14% of security professionals allow AI to autonomously execute remediation actions without human oversight — meaning most enterprises, despite adopting AI tools, still have their defense speed limited by human decision-making. Meanwhile, attackers' AI operates at machine speed.

Platform Consolidation Is the Only Way: 93% Prefer Full Platforms Over Point Products

In line with last year's findings, 93% of security leaders prefer buying comprehensive platforms over individual point products when acquiring new cybersecurity capabilities. Fewer vendors mean tighter integrations, fewer console switches, streamlined management, and stronger cross-domain threat insights.

The Only Solution: Let AI Defense Be Faster Than Humans

The core contradiction revealed by this report is stark: attackers use AI to strike at machine speed, while defenders use AI but defend at human pace. This speed gap is the single largest vulnerability in today's security landscape.

To truly close this gap, organizations must achieve "autonomous AI defense" — not by removing humans entirely, but by freeing them from repetitive tasks so they can focus on strategic decisions, while AI automatically handles threat detection, blocking, and remediation at the edge. This is precisely the core value of 100% on-premise AI models + AIOps automation.