Cybersecurity
Airlock Digital Launches Agentic Steering to Control AI Agent Actions Before Execution

Key PointsAirlock Digital evaluates AI agent actions before execution rather than trusting the…
View more
- Airlock Digital evaluates AI agent actions before execution rather than trusting the agent's application wholesale.
- Denials return explanations and approved alternatives so agents can retry within policy instead of improvising.
- Agentic Steering launches for Codex and Claude in version 7.1 of the endpoint security platform.
Airlock Digital has introduced Agentic Steering in version 7.1 of its endpoint security platform, giving organizations a way to evaluate actions from supported AI agents before they execute and return policy decisions directly to the agent. The capability launches for Codex and Claude.
The change addresses a gap that appears when an approved AI tool starts operating other software. Trusting the agent’s application does not necessarily mean every command it generates should be allowed. Airlock Digital lets administrators define those boundaries, stop prohibited actions, and supply an explanation or an approved alternative so the agent can continue within policy.
From Trusting an Application to Governing Its Actions
Traditional application control determines which software may run on an endpoint. AI agents add another layer: once started, they can invoke applications, work with files, and revise their approach as they pursue a task. On a shared workstation, a person and an agent may otherwise inherit access to many of the same resources.
Airlock Digital’s Agentic Control product page describes separate centrally managed policies for people and agents on the same endpoint. Agentic Steering evaluates proposed actions from supported agents against the organization’s rules, permits authorized actions, and stops those that are prohibited.
That is a more specific decision than whether the agent itself belongs on an approved-software list. A company may authorize an assistant for development work while restricting particular commands or administrative operations it attempts during that work. The rules need to follow the actor and action, rather than treating every operation from a trusted application as equally acceptable.
Returning a Useful Decision to the Agent
A conventional block notification may tell a person what went wrong without giving an autonomous tool enough information to choose a safe next step. An agent can interpret failure as a reason to try a different route toward its original objective.
Agentic Steering returns the decision directly to the supported agent. Administrators can attach context explaining why an action was denied or identify an alternative that the organization permits. The intended result is a policy-aware retry, rather than a sequence of increasingly improvised attempts.
For example, an organization could configure guidance pointing an agent toward an approved tool when a requested command is blocked. The enforcement decision remains outside the agent: an explanation helps it plan what to do next, but does not grant permission to bypass the rule.
“Traditional application control determines what software is trusted to execute. Agentic AI introduces a new actor on the endpoint, with an ability to execute other applications,” said David Cottingham, Airlock Digital’s co-founder and chief product officer. “Teams now need the ability to define and enforce what an AI agent is allowed to do once it’s running.”
Discovery, Enforcement, and Session Review
The release builds on the AI discovery capabilities introduced in Airlock Digital 7.0. Security teams can identify AI agents, language-model tooling, and harnesses across endpoints, then use deny-by-default application control to restrict execution to approved software.
Agentic Steering adds action-level evaluation for supported agents. Together, those capabilities separate two questions: whether an AI application should be running at all, and whether its proposed behavior is permitted once it is running.
Airlock Digital also centrally records agent sessions, commands, files, policy decisions, outcomes, risk activity, token usage, and estimated cost. Teams can search sessions and review AI-generated summaries to understand what an agent attempted and how enforcement affected the result.
Those records can inform policy changes. Frequent attempts to use an unapproved tool might reveal an unclear workflow or a missing approved alternative. Reviewing subsequent actions can also show whether the agent used the guidance supplied after a denial. This feedback helps administrators refine operational rules; it does not establish that the product retrains an underlying AI model.
How It Extends Airlock Digital’s Existing Platform
Airlock Digital’s application allowlisting technology controls execution of applications, scripts, and processes. Its published features include trust rules at the file, path, publisher, and parent-process levels, alongside exception workflows. Agentic Steering applies that preventative security approach to a new source of endpoint activity.
The wider platform supports Windows, macOS, and Linux and integrates with existing security and IT workflows. That broader platform coverage should be distinguished from the new capability’s stated agent support: the launch specifically identifies Codex and Claude, rather than promising identical control over every AI tool.
Airlock Digital positions endpoint execution control as complementary to identity security, data protection, AI platform controls, and detection and response. Each addresses a different part of an agent’s operation. Checking whether a command may execute does not by itself establish that every model response is accurate or that every use of data is appropriate.
Availability and the Enterprise Implication
Agentic Steering is available to Airlock Digital customers with version 7.1 for Codex and Claude. The capability was previewed at Black Hat USA 2026 under the name Agentic AI Control & Governance.
The launch makes a practical distinction for enterprises adopting autonomous tools: approving an agent is the beginning of access control, not its final decision. By connecting execution policy with guidance the agent can use, Airlock Digital is trying to make security boundaries part of the working loop between an objective, a proposed action, and its result.












