No Priors
Building an AI Guardian for Enterprise with Onyx Security CEO Maxim Bar Kogan
Why It Matters
As AI agents become integral to critical business processes, the potential for unintended or malicious actions grows exponentially, threatening data integrity and operational continuity. Enterprises need a practical, scalable way to ensure these agents act safely, making Onyx’s AI‑guardian approach a timely answer to a rapidly emerging security gap.
Key Takeaways
- •Autonomous AI agents cause exponential security risks for enterprises
- •Onyx builds a control plane overseeing AI agent actions
- •Over 50% of enterprise AI usage now autonomous coding agents
- •Traditional identity and endpoint security insufficient for AI governance
- •Small purpose-built models enable low‑cost, low‑latency AI oversight
Pulse Analysis
The rapid adoption of autonomous large‑language‑model agents has turned a promising productivity tool into a security liability. Recent incidents—agents unintentionally publishing source code, leaking tokens, or even deleting databases—show that the volume of AI‑driven actions can grow exponentially, outpacing any human‑in‑the‑loop review. Enterprises that once worried only about data leakage into chatbots now face market‑wide panic as these agents take on critical infrastructure tasks. Without a way to validate each autonomous decision, the risk of catastrophic mis‑behaviour becomes unacceptable, prompting a surge in demand for dedicated AI governance solutions.
Onyx Security answers that demand with a dedicated control plane that monitors and validates every AI‑agent action across an organization. The company classifies enterprise AI into three buckets: low‑code automations, first‑party custom agents, and fully autonomous coding assistants—today, more than half of AI activity falls into the latter category. Traditional identity‑based permissions and endpoint protections cannot keep pace because agents need broad access to be useful, yet that very access makes them blind spots for existing tools. Onyx trains compact, purpose‑built models that act as cheap, low‑latency sentinels, flagging only high‑risk operations for deeper review, thereby preserving performance while reducing exposure.
The venture is powered by Israel’s renowned cyber‑security talent, many of whom come from elite intelligence units and bring deep expertise in mathematics, mechanistic interpretability, and AI infrastructure. This blend of cyber and AI research positions Onyx to tackle the long‑term challenge of controlling advanced models that could underpin $10 trillion‑scale AI companies. Analysts estimate a $100 billion‑plus market for enterprise AI‑governance platforms, and Onyx’s early foothold gives it a strategic advantage. As autonomous agents become ubiquitous, businesses that adopt a proactive oversight layer will avoid costly outages and safeguard their digital assets.
Episode Description
We are now closer than ever before to living in a world where AI agents are smart enough to run our power grids and manage water supplies. How do we keep them from going rogue? Sarah Guo sits down with Maxim Bar Kogan, founder and CEO of Onyx Securities, to explore the complexities of supervising and securing autonomous agents at the enterprise level. Maxim explains Onyx’s product as an AI control plane, which oversees the permissions and flexible contexts of agents while balancing latency, cost, and reliability. He also discusses how current controls have insufficient context to monitor agent intent, tradeoffs for gradual model rollout, the need for vendor-independent oversight, and Israel’s growing AI and security talent ecosystem. Plus, why Maxim is all-in on AGI.
Sign up for new podcasts every week. Email feedback to show@no-priors.com
Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @maximbarkogan
Chapters:
00:00 – Cold Open
00:45 – Maxim Bar Kogan Introduction
01:10 – AutoGPT and Betting on Agent Actions
05:17 – What Onyx Product Does
07:47 – State of Deployment in Large Enterprises
09:58 – Securing Agents
12:45 – Why Proxies Don’t Work
14:11 – Why Onyx Trains Its Own Models
18:38 – Onyx’s Talent Culture
21:24 – Mechanistic Interpretability
23:35 – How Onyx Builds Customer Trust
25:10 – Mitigating Risk at the Foundational Level
27:45 – Phased Rollout of Glasswing and Daybreak
29:11 – Large Enterprise Holdouts
30:46 – Onyx and the Larger AI Security Space
32:36 – Should Labs Address Model Trust and Governance?
36:56 – What Needs to Happen in Security
39:14 – Why Maxim is AGI-Pilled
41:15 – Conclusion
Comments
Want to join the conversation?
Loading comments...