Delinea and Neo both announced platforms this week that evaluate individual AI agent actions against enterprise policy before execution, targeting the security gap between authentication and behavior. Delinea is extending its privileged access management platform with runtime controls built on its StrongDM acquisition from earlier this year, while Neo emerged from stealth with $100 million in funding to build what it calls “Agentic Software Control.”
Delinea: Policy Enforcement at the Action Layer
Delinea’s update lets its platform analyze tool calls, database queries, SSH commands, and interactions with Kubernetes clusters and cloud environments before an agent executes them. Each requested action is evaluated against enterprise policy and can be approved, blocked, or escalated for human review.
“The security industry spent years solving credential theft, but AI agents have introduced a new problem: authorized access doing unauthorized things,” Delinea CEO Art Gilliland told Biometric Update.
The platform injects credentials just-in-time when access is required, scopes them to the specific task, and revokes them on completion. An agent might need read access to a production database to retrieve records, but that does not mean it should be able to delete data, modify permissions, or export unrestricted datasets. Delinea’s runtime layer enforces that distinction per action, per session.
Neo: $100M to Map the Agent Attack Surface
Neo took a different entry point. The company, founded by former SentinelOne COO Nick Warner, SentinelOne detection engineering lead Shlomi Salem, and former IDF Unit 8200 vulnerability researcher Eran Shirazi, launched with $100 million from Andreessen Horowitz and Bessemer Venture Partners, with participation from Craft Ventures and Merlin Ventures, according to Neo’s announcement.
Where Delinea starts from privileged access management, Neo starts from visibility. The platform inventories AI agents, AI-enabled applications, browsers, developer tools, SaaS platforms, plugins, extensions, and MCP servers across an enterprise, then analyzes what each application can do, what systems it can access, and what risks those capabilities create.
“Organizations need visibility into what this software can do and the ability to govern its behavior in real time,” said Zane Lackey, general partner at Andreessen Horowitz, in Neo’s launch announcement.
Neo attributes each action to the responsible human, agent, application, or identity, then enforces policies on tool calls, API access, data movement, and agentic workflows. The company cited Gartner research indicating that only 5 percent of enterprise applications featured agentic capabilities in 2025, but 40 percent will by the end of 2026.
A Category Taking Shape
The two launches represent converging approaches to the same architectural problem. Traditional identity and access management verifies who an agent is and grants access. Runtime authorization evaluates what the agent does with that access, action by action.
Delinea and Neo are not alone. Biometric Update reported that Nuggets recently introduced an Authority Control Plane that evaluates agent identity, delegated authority, organizational policy, and runtime context before approving actions. Lasso launched an AI Security Posture Management platform earlier this month that continuously maps agent ecosystems and detects runtime behavior drift.
The timing tracks with real incidents. OpenAI disclosed last week that a rogue AI agent compromised four third-party service accounts during a 4.5-day autonomous campaign involving 17,600 logged actions. Runtime authorization would not have prevented the initial breach, but per-action policy enforcement could have blocked the agent from escalating its access across systems after authentication.
For enterprise security teams deploying autonomous agents into production, the question is shifting from “how do we verify this agent’s identity?” to “how do we control what this agent does next?”