Field noteNov 20, 20265 min read1 source

Airlock Digital v7.1: Native Application Control for Your Agentic AI Workforce

In one minute

Airlock Digital’s v7.1 announcement adds agent-specific endpoint policies for supported AI agents, checking proposed commands before they run and returning configured guidance when a command is denied. For organizations using the product, this creates a distinct control layer for agent actions rather than treating an agent exactly like its human user.

What happened

Airlock Digital announced version 7.1 on November 17, 2026. The company describes the release as adding native application-control capabilities for agentic AI and making application-control policy faster to build and enforce. Existing customers are scheduled to receive the release beginning November 30; support-led upgrades will be arranged in waves, with Support confirming timing for each environment. Prospective customers can contact Airlock Digital from November 17 to arrange a demonstration or discuss a proof of concept.

The main feature, Agentic Control, applies policy separately to supported AI agents and the people using the same endpoint. Airlock Digital says it checks each command an agent proposes before execution. Permitted commands proceed; prohibited commands are denied. The company’s rationale is that approving an agent to run does not determine what it may do once running. Under this model, an organization can allow an agent to operate while restricting particular commands or actions without changing the user’s policy.

Agentic Steering adds feedback when a command is denied. Administrators can configure an explanation, identify an approved alternative, direct the agent to request human authorization, or tell it to stop. The guidance does not alter the policy decision or grant permission: every subsequent command is checked again, and switching tools does not make a prohibited operation permissible. Airlock Digital’s example is a coding agent whose package-install command is blocked; the configured feedback points it toward an approved package source, after which the replacement command is also checked.

The release also groups supported-agent activity into sessions. Administrators can review attempted commands, policy decisions and outcomes to see whether an agent followed an alternative, continued pursuing a prohibited operation, or stopped. The company says this history can help teams investigate repeated denials, determine whether an approved alternative failed because the agent lacked a needed tool or access, and confirm that prohibited actions were stopped while approved work was completed.

Version 7.1 also extends Trust Builder, introduced in version 7.0. It adds hash rules and combined conditions alongside publisher, path and process rules; one example combines publisher and path to restrict trust to a location while allowing legitimate updates. Recommendations may be reviewed individually or automatically applied when eligible under selected settings. Files identified as suspicious or malicious by Airlock Digital Intelligence, powered by VirusTotal, are held for manual review rather than automatically applied. Optional AI explanations can be generated by connecting an organization’s own AI provider. Airlock Digital says recommendations analyze the organization’s own environment, not other customers’ data, and that policy change history records rules Trust Builder adds.

What the archive adds

New Runtime’s earlier coverage of enterprise controls described PwC’s view that inventories and controls should cover the full transaction chain and orchestration layer. Airlock Digital’s announcement is a new product event, not a restatement: it specifies endpoint-level command checks and feedback for supported agents. The earlier coverage is a broader governance argument, rather than reporting on this release.

Trend signal

This release supports agent security moving to runtime boundaries. Airlock Digital places enforcement at the endpoint: policy is checked before an agent’s proposed command executes, and a denial does not depend on the agent simply following a prompt. The company’s announcement describes a vendor capability, not independent evidence that endpoint controls alone are sufficient for agent security.

Why now

Airlock Digital frames agents as a new endpoint actor that can choose tools, run commands and execute applications while working toward a task. Its stated timing is the need to adopt agents without giving up control or slowing work. The company announced v7.1 on November 17, with existing-customer availability scheduled to begin November 30.

Why it matters

For teams already using Airlock Digital, the distinction between permission to run an agent and permission for that agent to execute particular commands is operationally important. A user’s access can be broader than an organization wants an autonomous process to exercise. Separate agent policy offers a way to constrain that process without changing the person’s rules, while per-command checks preserve enforcement when the agent changes its approach.

The feedback loop is also consequential: a denial can point toward a permitted route, ask for human authorization, or end the attempt. That could reduce unproductive retries while leaving policy authority with administrators. Session history gives security teams a record to inspect when an agent does not follow guidance or a permitted alternative does not work. These are useful control primitives, but their value depends on which agents are supported, how policies are configured and whether teams can review the resulting activity. The announcement does not establish outcomes across customer deployments.

What happens next

  • Existing Airlock Digital customers will begin receiving v7.1 through the announced upgrade waves (3 months, 90%). Watch for: the company’s support-led rollout beginning November 30 and environment-specific scheduling.

Limits and uncertainty

Airlock Digital says command checks apply to supported AI agents, but its announcement does not quantify the supported-agent set or describe coverage across agent versions and endpoint operating environments. It provides no customer deployment measurements for blocked actions, successful redirections, task completion, or administrative effort. The announcement also does not independently establish how the controls behave under adversarial attempts or how quickly policies must be maintained as agents and tools change. Trust Builder’s optional AI explanations are described as explanations of recommendation scope and tradeoffs; the company does not claim they determine or enforce policy.

Reader takeaway

Airlock Digital v7.1 separates agent permissions from user permissions and checks each supported agent command against endpoint policy. For organizations evaluating it, the key questions are whether their agents are supported, how denials and approved alternatives are configured, and whether session history gives security teams enough evidence to validate the controls in practice.

Evidence / sources

  1. 1airlockdigital.com/airlock-blog/v7.1-native-application-control-for-your-agentic-ai-workforce
All Field Notes

Next reads

Read next

Open the graph
  1. 01topicAgent SecurityExplore the agent-security topic hub.
  2. 02topicSecurityExplore the security topic hub.
  3. 03archiveField NotesOpen the latest editorial analysis.
  4. 04sourcesSourcesSee the original sources this archive cites.
Who read this page?Machine requests, hidden until opened

Loading the privacy-safe route aggregate...

Open the JSON contract