# Through the AI Fog: The Architectural Decision Agentic Security Depends On — Manoj Nair, Snyk

## Executive summary

As autonomous agents and frontier LLMs accelerate development speed, they simultaneously create a novel and expanding attack surface. The core security challenge is that probabilistic systems (like large models) cannot be trusted to police themselves. Data shows significant growth in the security backlog (108% quarter over quarter). To build safe, agentic software at scale, organizations must move beyond relying solely on model intelligence and implement deterministic verification layers that validate agent output, skills, and environment interactions.

## Key takeaways

- The Generator vs. Validator Problem: A fundamental security principle is questioned: Can the system generating code (the generator) also be the system verifying it (the validator)? The answer, according to real-world data, is no.
- Exponential Vulnerability Growth: The security backlog for customers grew by 108% quarter over quarter (QoQ), indicating that the rate of vulnerability creation is outpacing remediation efforts.
- New Attack Vectors in Agentic Systems: Threats include 'toxic skills' (where a third or more of all available skills contain malware), insecure connections via MCP servers, and agents quietly copying PII into untrusted databases.
- Deterministic Verification is Essential: When testing for vulnerabilities, the latest frontier models found only 75% of issues in red team attacks, compared to a deterministic checker which achieved at least a 40% F1 score. This highlights that probabilistic systems require supplementary validation.

## Technical details

- Agentic Code Risk: The risk is not limited to model output; agents introduce an entirely new threat surface. For every model found in a repository, there are three times more agentic components (in tools and skills) that must be assessed.
- Vulnerability Chaining: Novel exploits often do not require novel vulnerabilities. Instead, they chain together existing low-severity issues to create a much larger blast radius.
- Security Tools and Remediation: The Snyk package health check tool is recommended for assessing open-source dependencies, determining if they are actively maintained (healthy) or merely vulnerability-free. The 'Evo' product offers agentic dev security by monitoring environment, output, skills, and MCP servers in real time.

## Practical implications

- Implement deterministic security checks alongside LLM outputs to validate code and system behavior.
- Treat agent skills and MCP servers as high-risk components that require explicit risk assessment for malware or hostile instructions.
- Adopt tools (like Snyk's package health check) that assess not just CVE status, but also active maintenance and usage patterns of dependencies.
- Focus governance efforts on the entire agentic loop—including environment changes, skill calls, and output handling—rather than just model inputs.

## Topics

AI Security, Agentic Development, LLM Governance, Vulnerability Management, Build Automation Security, Snyk, Evo

Source: https://www.youtube.com/watch?v=1EZdpEhwmNc