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

AI agents have become a new actor on the endpoint, and enterprises want to adopt them without giving up control or slowing the business down.

Airlock Digital v7.1 is built for both.

With Agentic Control, you decide what AI agents can do on your endpoints, just as you decide what software can run. Enhanced Trust Builder, introduced in version 7.0, helps you build that policy and reach enforcement faster, with new rule types, review safeguards, and optional AI explanations.

Introduction

At Airlock Digital, we’ve spent more than a decade helping enterprises put application control to work. Our Deny by Default approach stops attackers before they can act, without disrupting the business.

Agentic AI has quickly become part of how modern organizations work, and it brings a new actor to the endpoint. AI agents share the same endpoints, applications, and files as the employees they work for. They choose tools, run commands, and execute other applications on their own as they work toward an objective. Without separate policies, an agent has the same access as the person using it.

Application control has always determined what software is trusted to run. Approving an AI tool lets the agent run, but it doesn’t define what the agent is allowed to do once it’s running. Enterprises now need to answer that second question too. They also need policy they can put in place quickly, without slowing down the work that makes agents valuable.

Today we’re announcing Airlock Digital v7.1, a significant release for our enterprise customers that does both. It brings native application control capabilities into your agentic AI workforce, and it makes the policy behind your application control faster to build and enforce. We’re excited to put it in your hands.

Here’s what’s new in v7.1:

  • Agentic Control. Set policy for supported AI agents separately from the people on the same endpoint and check every agent command against that policy before it runs.
  • Agentic Steering. Don’t just block. When a command is denied, Airlock Digital tells the agent why and points it to an approved alternative. The agent can keep working within policy.
  • Enhanced Trust Builder. Building on its 7.0 release, Trust Builder adds new rule types, review safeguards, and optional AI explanations to help you build application control policy and reach enforcement faster.

Agentic Control Decides What AI Agents Can Do

An organization approves agentic AI tooling to help its employees work. An employee gives an AI agent a legitimate task, and the agent starts choosing tools and issuing commands to complete it. The agent may interpret the request too broadly, choose a method the organization does not permit, or keep trying different commands after a denial.

Agentic Control applies application control to those actions. Approving an agent lets it run on the endpoint, and your policy decides which commands and actions it can carry out. Because agent policy is separate from user policy, you can restrict what an agent does without changing the rules for the people on the same endpoint.

Agentic Steering Guides the Agent’s Next Step

Blocking a command is only part of the picture. Agents revise their approach when a step fails, and blocked command may lead an agent to try another tool or method. Even when every unapproved execution is stopped, repeated attempts can consume time, tokens, and resources without completing the task. The agent needs to understand what the restriction means for its next step.

Agentic Steering, a capability within Agentic Control, combines enforcement with policy feedback and guidance. Airlock Digital checks each command a supported agent proposes before it runs. Permitted commands go ahead. Prohibited commands are denied, and the agent receives the decision. With each denial, administrators can:

  • Explain the restriction.
  • Identify an approved alternative.
  • Instruct the agent to ask a human for authorization, or stop.

Guidance helps the agent respond to policy, but it does not change what policy allows. Every subsequent attempted command is still checked. Switching tools does not grant permission for a prohibited operation, and an instruction to request approval does not grant that approval.

Guidance Keeps Approved Work Moving

For example, an approved coding agent tries to install a package in a way the organization’s security policy does not allow. Airlock Digital denies the command before it runs. Using feedback the team configured, it directs the agent to the approved package source. The agent proposes a different command, and Airlock Digital checks that one against policy too.

Security teams can define a permitted way forward and see whether the agent takes it. Where there is no permitted alternative, the guidance can tell the agent to ask a human for authorization or stop the command from executing. Teams can hand work to agents and still decide which endpoint actions those agents can run.

Session History Helps You Refine Policy

After a policy decision, teams can see what happened next. Airlock Digital groups supported AI agent activity into sessions that show each command an agent attempted, the policy decision, and the outcome. Teams can see whether the agent followed an approved alternative, kept pursuing a prohibited operation, or stopped.

Teams can use this history to:

  • Check the explanation an agent received if it repeatedly attempts a denied command.
  • Investigate whether the agent has the tool or access it needs if an approved alternative fails.
  • Confirm that prohibited actions were stopped and that the agent completed its approved work.

 

Trust Builder Helps You Build Application Control Policy Faster

Since its introduction in version 7.0, Trust Builder has recommended trust rules based on your organization’s execution history and the security preferences you configure. Teams can:

  • Define eligible rule types and combined conditions. For example, a team that prefers publisher trust limited to specific paths can set that preference and check recommendations against observed activity.
  • Review evidence and alternatives. See the evidence behind each recommendation, compare alternatives, and choose where to apply rules.
  • Go from investigation to rule. Move straight from investigating a file to applying a rule for it.
  • Choose how much to automate. Review each recommendation before accepting it or let Trust Builder apply eligible rules automatically within the policies and settings you select. Evidence and change history show what was applied and why.

 

New in v7.1

  • More rule types: Trust Builder now recommends hash rules and combined conditions alongside publisher, path, and process rules. For example, combining publisher and path limits trust to a specific location while still allowing legitimate updates.
  • Airlock Digital Intelligence powered by VirusTotal: recommendations that include files Airlock Digital Intelligence identifies as suspicious or malicious are held for manual review instead of being applied automatically.
  • Optional AI explanations: connect your own AI provider to explain a recommendation’s scope and tradeoffs, so your team can review it faster. Recommendations come from analysis of your own environment, not other customers’ data.
  • Review or automate, per policy: approve each recommendation yourself, or let Trust Builder apply eligible rules automatically within the settings you choose. Policy change history shows every rule Trust Builder adds.
  • From investigation to rule: compare trust methods for a file, see related files and coverage, then add a rule directly to a policy group or existing allowlist.

Availability and Next Steps

Airlock Digital v7.1 will be available to existing customers starting on November 30th. To have Airlock Digital Support perform your upgrade, contact the team now. Support-led upgrades will be scheduled in waves from November 30th, and Support will confirm timing for each environment. If you are evaluating Airlock Digital, contact our team to arrange a demonstration or discuss a new proof of concept starting on November 17th.

Book a meeting with Airlock Digital to explore Agentic Control with Agentic Steering in version 7.1 and see how enhanced Trust Builder can simplify application control policy creation and management.