Building an AI Guardian for Enterprise with Onyx Security CEO Maxim Bar Kogan

2026-05-28 · Source: No Priors: AI, Machine Learning, Tech, & Startups · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Intermediate, extended

Summary

Onyx Security, led by CEO Maxim Bar Kogan, addresses the escalating security risks posed by the exponential adoption of AI agents in enterprises. With over 50% of current enterprise AI deployments being autonomous coding agents like Claude Code and OpenClaw, organizations face challenges such as accidental code publishing, token leaks, and data deletion. Traditional security measures, including identity and endpoint protection, are insufficient as they lack the context to understand agent intent or cannot be overly restrictive without hindering productivity. Onyx's solution involves a "secure control plane" that trains small, specialized AI models to efficiently oversee other agents. These guardian models identify high-risk actions, triggering intervention by smarter agents or humans, thereby balancing performance, cost, and security. This approach is crucial as the cost of vulnerability finding plummets, necessitating foundational AI security.

Key takeaway

For Directors of AI/ML and CTOs rapidly deploying autonomous AI agents, recognize that traditional security paradigms are inadequate for managing the exponential risks of agent actions. Your teams must invest in specialized, AI-native security solutions that provide independent oversight, such as a secure control plane. This approach allows for the safe scaling of agent adoption by efficiently identifying and mitigating high-risk behaviors, preserving productivity while fortifying your enterprise against emerging threats like plummeting vulnerability finding costs.

Key insights

Enterprises need independent AI guardians to oversee autonomous agents, as traditional security and vendor-provided controls are insufficient for exponential risk.

Principles

• AI agent actions grow exponentially, outpacing human oversight capacity.
• Traditional security tools lack context for autonomous AI agent intent.
• Specialized, small AI models can efficiently detect high-risk agent actions.

Method

Train small, focused AI models to act as a "secure control plane," monitoring other agents and flagging critical actions for review by smarter agents or humans, balancing performance and risk.

In practice

• Implement a secure control plane for autonomous coding agents like Claude Code.
• Prioritize foundational AI security tools to mitigate Mythos-level vulnerability risks.
• Adopt a nuanced AI adoption strategy based on enterprise risk profile.

Topics

• AI Agent Security
• Autonomous Agents
• Enterprise AI Adoption
• AI Governance
• Mechanistic Interpretability
• Onyx Security
• Vulnerability Finding

Best for: VP of Engineering/Data, Investor, Executive, AI Security Engineer, Director of AI/ML, CTO

Related on AIssential

• Securing AI agents with agent governance — Applying operating system and distributed systems security principles is crucial for governing autonomous AI agents.
• How enterprise leaders are scaling AI agents across their organization — Scaling AI agents responsibly requires integrating governance, managing complex workflows, and fostering controlled experimentation to build trust and deliver value.
• Govern your bots carefully or chaos could ensue — Effective AI agent governance, not restricted access, drives higher value and mitigates "agent sprawl" risks.
• The Pentagon Threatens Anthropic — Autonomous AI agents necessitate a shift from intent-based trust to structural safety architecture across all interaction levels.
• Mastering Agentic AI: Governance, Trust, and Enterprise Success with Barndoor AI CEO and Founder Oren Michels — Effective agentic AI adoption in enterprises requires robust governance and fine-grained access control to build trust and manage risks.
• Posthuman: We All Built Agents. Nobody Built HR. — Effective enterprise agentic AI requires robust, out-of-band governance infrastructure, not just better models.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by No Priors: AI, Machine Learning, Tech, & Startups.