$113 million Series B led by Bessemer Venture Partners<\/a>, bringing total funding to $153 million in two years. Its platform sits between AI agents and corporate infrastructure, monitoring each step of an agent’s reasoning to catch policy violations before they hit production. Customers include Fortune 500 firms in banking, insurance, energy, and technology.<\/p>\nWhat this means for your business<\/h2>\n
The question Onyx is answering is one most security teams haven’t formally assigned to anyone yet: who owns the audit trail when an AI agent, not a human, executes a multi-step transaction inside your core systems? If your organization has deployed ChatGPT Enterprise or Claude Enterprise in any workflow that touches regulated data or operational infrastructure, you already have this exposure. The companies that haven’t mapped their agent footprint are the ones this funding round should make uncomfortable.<\/p>\n
Onyx’s core claim is that existing security tooling, built for human-initiated actions, breaks down when agents execute hundreds of decisions autonomously with limited checkpoints. That’s structurally correct. Traditional identity and access management assumes a person authenticates and acts; an agent authenticates once and then acts continuously, making the point of intervention the behavior itself, not the login. Onyx’s approach of analyzing agent reasoning at each step rather than just at the boundary is the right architectural instinct, though the practical question is whether its proprietary models can keep pace with the speed and variety of agent actions at enterprise scale without becoming a latency bottleneck.<\/p>\n
Bessemer’s Amit Karp framing this as a trust and accountability problem, rather than a pure threat-detection one, is the tell. The firms writing the biggest checks into this category believe governance will be a procurement requirement before it’s a solved technical problem, which means the window for CISOs to define their own standards is closing. If your AI governance policy today still routes through your general AI acceptable-use document rather than a purpose-built agent oversight framework, the renewal conversation with your board is going to raise that gap before your vendor does.<\/p>\n
Concept deep-dive: Agent control layer<\/h2>\n
An agent control layer is software that sits between autonomous AI agents and the enterprise systems they’re allowed to act on, functioning like a security checkpoint that watches not just whether an agent has permission to enter a system, but whether what it’s doing inside that system matches expected behavior. Think of it as the difference between checking a contractor’s badge at the door versus monitoring what they actually do once they’re on the floor. The business case is accountability without slowing deployment.<\/p>\n
Based on reporting from Onyx Security Raises $113M Series B to Expand Enterprise AI Governance<\/a>, originally published 2026-08-03 05:50:00.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"Share with your CISO Onyx Security is betting that the shift from humans to AI agents as the primary actors inside enterprise systems creates a security gap large enough to justify a dedicated control layer. The New York startup closed a $113 million Series B led by Bessemer Venture Partners, bringing total funding to $153 […]<\/p>\n","protected":false},"author":1,"featured_media":7577,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[143],"tags":[238],"tmauthors":[],"class_list":["post-7576","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai-security","tag-ciso"],"_links":{"self":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts\/7576","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/comments?post=7576"}],"version-history":[{"count":0,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/posts\/7576\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/media\/7577"}],"wp:attachment":[{"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/media?parent=7576"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/categories?post=7576"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/tags?post=7576"},{"taxonomy":"tmauthors","embeddable":true,"href":"https:\/\/workai.tv\/news\/wp-json\/wp\/v2\/tmauthors?post=7576"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}