IDENTOS Launches PolicyArc: A Permission System Built for AI Agents

AI agents are moving from demos into real enterprise workflows, and with them comes an unsolved problem: how much access should a piece of software that acts on your behalf actually get? Traditional permission systems were built for people who log in and out, not for software that keeps working while you are away. This week, the Canadian identity and access company IDENTOS announced PolicyArc, a new platform built to answer exactly that question.

The timing is what makes this matter. Enterprises are wiring AI agents into internal tools that pull customer data, trigger actions, and make decisions once reserved for humans. Security teams are being asked to say yes without a clear way to limit what the agent can touch. PolicyArc is IDENTOS arguing that agents should never get the broad, standing access people get.

The Problem With Handing Agents the Keys

The way most companies handle access today is simple. An employee joins, gets a role, and that role comes with a bundle of permissions that persist until somebody removes them. It works well enough for people, because people are roughly accountable for their own behavior.

Agents break that model. An AI agent acting on your behalf might need to read a handful of customer records or run one report. Hand it your full permissions, and you have given software that can run thousands of actions per minute the keys to everything you can see. A malicious prompt or simple confusion can turn that into damage at speed and scale.

IDENTOS says this is the gap PolicyArc was built to close: agents are a third category of actor that needs its own rules, narrow, temporary, and constantly rechecked.

What PolicyArc Actually Does

PolicyArc is a policy-based access control platform. It sits between the AI agents (and people and services) on one side and the applications they need to reach on the other, and it decides what each one is allowed to do.

The headline feature is scoped delegation. A person can grant an AI agent a limited version of their own access, just enough for the task at hand, for a defined window of time. When the task finishes or the window closes, the access disappears automatically. There is no lingering permission waiting to be forgotten months later, which is how most access sprawl starts.

Next is consent enforcement. Normal access systems check permissions at login and then trust the session. PolicyArc checks each request against the policy and what the data owner has actually agreed to. That matters for sensitive data, where consent can be conditional rather than a one-time yes.

Then there is real-time decision making. Instead of relying only on standing facts like job role or verified identity, PolicyArc folds in live context: how sensitive is the resource, what stage is the workflow in, does anything look off. Based on that, it can allow the action, deny it, or ask for extra verification.

One policy applies across every connected application, and every decision is logged by default, so audit evidence exists from day one instead of being reconstructed after an incident.

Why This Launch Is Timed to the AI Breach Numbers

IDENTOS is framing PolicyArc around a striking statistic. The company says that, per IBM and Ponemon Institute's 2026 Cost of a Data Breach Report, 92 percent of organizations that experienced an AI-related breach lacked proper AI access controls. The implication is blunt: companies are adopting agents faster than they are governing them.

"Most systems only know how to grant broad, standing access. That is fine for a person, but reckless for an agent," said Alec Laws, chief technology officer at IDENTOS, in the launch announcement. "We built PolicyArc so a team can hand an agent exactly the permission it needs for one task, nothing more. When the task ends, the access disappears, and every decision is on record."

Mike Cook, the company's CEO, called the launch "a natural evolution for IDENTOS in the era of AI," framing it as applying the company's identity and access experience to a question every enterprise now asks: how to adopt AI without losing control.

The Canadian Angle: A Decade in Regulated Sectors

PolicyArc did not come out of nowhere. IDENTOS was founded in 2014 and has spent more than a decade on identity and access across Canadian government and healthcare, two sectors where getting access wrong carries serious consequences.

That background shows up in the positioning. IDENTOS is ISO 27001 and SOC 2 Type II certified, and the company states that all of its products are built on 100 percent Canadian data residency. For Canadian enterprises and public sector organizations dealing with sovereignty requirements, that is a meaningful differentiator against larger US-based identity platforms.

The target audience is the people currently losing sleep over AI adoption: platform and security leads, compliance and risk teams, and product teams shipping AI-agent features that need safe delegation without building the infrastructure.

What It Costs and Where You Can Get It

PolicyArc is generally available now, and interested teams can find it at policyarc.com. The company has not published pricing in its announcement, which is typical for enterprise platforms, where costs usually depend on scale and deployment shape.

PolicyArc does not stop agents from making mistakes; it limits the blast radius when they do. Access control is one layer of AI security, sitting alongside prompt-injection defenses, output validation, and monitoring. IDENTOS is not claiming to solve AI security end to end.

The timing is sharp. As agents move from side experiments into systems that touch money, health records, and government data, what they are allowed to do stops being theoretical. IDENTOS is betting the answer is least-privilege access, enforced continuously, with a paper trail.

Frequently Asked Questions

What is PolicyArc?

PolicyArc is an access control platform from Canadian firm IDENTOS, launched October 20, 2026. It lets enterprises define, enforce, and audit access policies across their applications, with a focus on safely giving AI agents scoped, time-limited permissions instead of broad, permanent access.

How is this different from regular access management?

Traditional systems grant standing role-based permissions, checked mainly at login. PolicyArc is designed around delegation to AI agents: permissions are narrow and temporary, rechecked against live context and consent on every request, and automatically revoked when the task or time window ends.

Who should care about this launch?

Security and platform leaders, compliance and risk teams, and any organization deploying AI agents into production. It is especially relevant for regulated sectors like healthcare and government, where audit trails and data residency matter most.