Cybersecurity pain points

What people report in Cybersecurity

Classic security operations and a layer of AI agent control share this field. The agent side leads on reach: Unsafe Agent Actions touches the most people among the pain points listed, 557: AI agent developers and operators whose agents make changes that cannot be undone, or that destroy things, in live systems with no real human sign-off, limits or audit trail. AI Agent Governance Gap and Unauthorized AI Agent Actions are close relatives, describing missing identity, permission and lifecycle controls and agents holding broad permissions. The operations side is about volume: Vulnerability Triage Overload and Security Alert Fatigue bury Security and AppSec teams and SOC and security teams in noisy findings. The field has 29 pain points in all.

Security Alert Fatigue is hit at least weekly by 90% of the people describing it, AI Agent Governance Gap by 86% and Vulnerability Triage Overload by 80%. Manual Compliance Evidence comes at 41% and Security Remediation Tracking at 26%. Cost reports are broad: 277 people say Unsafe Agent Actions costs them and 269 say the same of Manual Compliance Evidence.

Claude, Claude Code, SIEM, MCP, Slack, EDR, Codex and Excel are the tools named most. Gaps differ by problem. For agents, no inspection layer sits between an orchestrator and its subagents, hooks block only part of what they should, and agents ignore rules and carry on regardless. A force-merge with admin rights went through despite branch protection and review of the pull request. For compliance, tooling turned paper forms into PDFs without automating the data, no one workspace brings together who did what, deadlines and collaboration, and papers sit in many places, which makes a legal claim slow to put together. Scanners list what exists but not whether it can be reached, lack business context and asset criticality, and their verdicts never become paperwork. For alerts, teams lack the people and hours to tune them or act on them, and success emails cannot be told from failures.

Manual Compliance Evidence is the strongest case in this list: 41 people pay or would pay for a fix, the most here, 269 say it costs them and 437 raised it. Vulnerability Triage Overload follows with 37 and the better frequency, 80% weekly against 41%. The AI agent problems draw large crowds but few would-be buyers: 557 people on Unsafe Agent Actions with 15 who would pay, and 263 on Unauthorized AI Agent Actions with 3.