AI Agents, Enterprise Scale, No Compromises: Now via AWS

A couple of years ago, AI agent security was a niche conversation. The practitioners who took it seriously were a small group of researchers, a handful of forward-looking CISOs, and a few founders who had watched the attack surface forming in real time. The broader market hadn't caught up yet.

It has now. Enterprises are deploying AI agents at scale across platforms. The productivity gains are real. The competitive pressure to adopt is real. And the space between how fast organizations are moving and how well they can govern what those agents actually do is very real.

That's why I'm proud to share that Zenity is now part of AWS Security Hub Extended. For the security community, this is a meaningful shift: AI agent risk has earned its place in the enterprise security stack. Not as a warning to slow down, but as the infrastructure that lets organizations move faster with confidence.

Why AI Agents Break the Security Model We've Relied On

For the past two decades, enterprise security has been built around a reasonably stable set of assumptions: protect the perimeter, secure the endpoints, govern identity, and monitor the network. These assumptions shaped the tools we built, the teams we organized, and the frameworks we trusted.

AI agents violate every one of those assumptions.

An AI agent doesn't just produce an output; it executes. It maintains state. It invokes tools. It chains actions across systems, crossing identity boundaries and cloud environments in a single workflow. When a Copilot Studio agent with access to your CRM, your SharePoint instance, and your production database receives a manipulated prompt, it doesn't just generate a bad response. It acts on it.

Traditional security tools were designed for a world where humans initiate actions and systems respond. They were not designed to govern autonomous execution initiated by an agent acting on behalf of a user, a workflow, or another agent. That gap isn't a product deficiency. It's a category mismatch.

Detection after exfiltration is not security. By the time traditional tools or point solutions flag an anomaly in agent behavior, the action has already been taken, the data has already moved, and the blast radius has already expanded. We need security that operates at the speed of the agent, before the harm is done.

The answer to that problem isn't to restrict what agents can do. It's to build the visibility and controls that make it safe for them to do more.

What 'Full-Stack' Actually Means Now

AWS Security Hub Extended was built around a simple but powerful insight: enterprise security is only as strong as its most underprotected domain. You can have world-class endpoint detection from CrowdStrike, identity governance from Okta, email security from Proofpoint, and network protection from Zscaler, and still be exposed if the AI agents operating across all of those domains are ungoverned.

That's the gap Zenity closes.

Within Security Hub Extended, Zenity's AI agent security findings integrate directly into the Security Hub risk correlation engine using the Open Cybersecurity Schema Framework (OCSF). That means AI agent posture violations, runtime anomalies, and detected threats appear alongside cloud, endpoint, and identity findings in a single pane of glass. No separate console, no custom integration, no data normalization overhead.

Consider what that looks like in practice: a phishing email detected by Proofpoint reaches an employee who uses an agent with access to production systems. That agent is targeted with a prompt injection attack, detected by Zenity's runtime threat detection. The agent attempts to exfiltrate data through an overprivileged tool integration surfaced first by continuous exposure and posture management, then blocked at execution by AI guardrails before any data moves. The correlation engine connects them into a single attack path, giving the security operations center a complete picture and a clear remediation path.

That is what full-stack security looks like in a world with AI agents in it. And critically: the agent kept working. The legitimate use case was never disrupted. Only the attack was stopped.

The Gap Between Configuration and Runtime

One of the things I've learned while building Zenity is that AI agent risk doesn't start at runtime. It starts earlier than most people think, when an agent is configured with excessive permissions, connected to sensitive knowledge bases it doesn't need access to, or deployed without the guardrails that would constrain its behavior under adversarial conditions.

By the time a misconfigured agent is running in production, the exposure is already baked in. Security teams are left reacting to incidents rather than preventing them. The goal is to continuously map every agent, its permissions, integrations, and data connections, and surface the risks that are actually exploitable before they reach runtime. Secure by design, not secure by accident.

Our native integration with Amazon Bedrock AgentCore is built on exactly this principle. Zenity provides full-lifecycle security coverage for organizations building agents on AWS, from continuous exposure management to guardrails that enforce what agents are allowed to do, at the moment of execution, across every turn and session. It's the same philosophy that shapes how we approach every platform we support: you can't govern what you can't see, and you can't protect what you haven't secured from the start.

The teams building the most ambitious agent deployments aren't slowing down for security. They're choosing security infrastructure that moves with them. That's what Zenity is built to be.

Why This Moment Matters

AWS Security Hub Extended represents more than a procurement convenience. It represents a structural shift in how enterprise security is assembled, governed, and operated. By unifying findings from across the security stack, including endpoint, identity, email, network, cloud, and now AI agents, it creates the conditions for security teams to finally operate with a complete picture.

For Zenity, joining this ecosystem reflects something we've believed from the beginning: AI agent security isn't a standalone category. It's a critical layer of the enterprise security stack that needs to interoperate with every other layer. The threats that involve AI agents don't stay contained to AI systems. They traverse identity, cloud, and data. The defenses need to match.

The partnership with AWS matters. But what it enables matters more: enterprises that can move fast with AI. Building agents, scaling them, and letting them act with the confidence that security is keeping pace. That's what we built Zenity for. And we're just getting started.

Check us out on AWS Security Hub Extended.