A Better Newspaper

## Overview A distinct enterprise security market is emerging around governing autonomous AI agents — software that can access tools, retain information, take actions across applications, and hand off tasks to other agents without waiting for human review. Multiple vendors and events in late September 2026 highlight this as a fast-consolidating category, with vendors now distinguishing between access/permissioning controls and "runtime governance" of agent behavior once deployed. ## Key Developments - At Okta's Oktane conference, analysts noted that agent security "now depends on knowing what agents can do after they gain access," since agents can act across applications and pass work to other agents faster than security teams can review changes (SiliconANGLE, September 24, 2026). - Palo Alto Networks Inc. is addressing the shift "from chatbots to autonomous agents," which changes enterprise risk since agents can access tools, retain information, and act without a human prompt (SiliconANGLE, September 24, 2026). - Dataiku Inc. launched Agent Management, a standalone product to find and track every AI agent a business runs, including those built on other vendors' platforms, aimed at closing a visibility gap at large companies (SiliconANGLE, September 24, 2026). - German startup Kontext raised $4 million for software that sits between an agent and the systems it acts on, checking every requested action against security policy before execution (SiliconANGLE, September 24, 2026). - Collibra BV introduced "runtime governance" capabilities for enterprise AI agents — Live Map, Maestro, Guardian Agents, and Agent Contracts — designed to govern both the business context agents receive and the operational boundaries they act within, framing agent oversight as extending beyond answering questions into taking action across enterprise systems (SiliconANGLE, September 25, 2026). ## Strategic Significance The core unresolved question raised in coverage is recovery: "How does a company recover" once an agent has acted incorrectly or maliciously (SiliconANGLE, September 24, 2026). This positions agent permissioning, visibility, and policy-enforcement tooling as a critical new enterprise infrastructure layer, with implications for liability, compliance, and incident response frameworks that attorneys and enterprise buyers will need to track closely. Collibra's entry suggests the category is bifurcating into pre-execution permissioning/policy tools (Kontext, Palo Alto Networks) and runtime/contextual governance layers (Collibra) that govern what business context and behavioral boundaries agents operate under during execution (SiliconANGLE, September 25, 2026). ## Related Entities Okta (Oktane), Palo Alto Networks, Dataiku, Kontext, Collibra BV, and numerous adjacent agent-security startups (Aviatrix, Zero Networks, SailPoint/Entro, Astrix Security, Xage Security, Cynomi) are all racing to define this space.