The underlying announcement was issued by Quest on September 16, then summarized by Security Today and republished in substantially the same form by Portal ERP on September 18. The two reports confirm the announced product expansion, but neither establishes independent deployment results, customer availability, licensing terms or effectiveness of the new AI-driven controls. Those omissions matter: one of the five announced capabilities, Secure Replay, is explicitly only in private preview, while Quest has not publicly said when the remaining functions will reach every existing customer or what product tiers they require.
The expansion targets the identity layer behind Microsoft environments
Quest’s new package spans five named capabilities: Identity Insights, Agentic AI Defense, Secure Replay, Guardian Managed Recovery Services and Secure Migration. Quest says the collection extends its platform across the National Institute of Standards and Technology Cybersecurity Framework lifecycle, from discovery through recovery.
Identity Insights is the foundation. The feature uses technology acquired through Quest’s June 2026 purchase of Anetac to continuously map what human, non-human and agentic identities can reach. That means the intended target is not simply a directory inventory. Quest says it wants to expose access paths that periodic scans and configuration reviews miss: a service principal with broad Microsoft Graph permissions, a stale account retaining group membership, or an agent connected to data sources through a long-lived credential.
That is a relevant problem in Microsoft estates, but it is not a new identity category in Entra itself. Microsoft defines workload identities as applications, service principals and managed identities, and its documentation now identifies agent identities as a particular form of machine identity requiring stronger governance. In practice, an “AI agent” may still authenticate using ordinary Entra constructs, OAuth permissions, a managed identity, a service principal certificate or a secret. Administrators should therefore judge Quest’s claims on whether it can accurately connect an agent’s business purpose and runtime activity back to those underlying Entra objects and permissions.
An inventory that merely adds an “AI” label to existing service principals would be of limited operational value. The potentially useful distinction is whether Quest can map delegation chains and effective access: who sponsored the agent, which identity it acts under, which APIs it can call, what consent and roles it inherited, and which systems it can alter.
Containment claims need a clear enforcement path
The most consequential claim in the release is Agentic AI Defense, which Quest says can autonomously isolate a compromised identity while an attack is still unfolding. Security Today reported the same description: the feature is intended to prevent further destructive actions while security staff investigate.
Quest has not publicly detailed what “isolate” means in each supported environment. That leaves several material questions unanswered for administrators: Does the product disable an AD account, revoke Entra sessions, disable a service principal, remove role assignments, alter Conditional Access policy, block a connector, or simply create a response recommendation for an analyst? Those are very different actions, carrying very different outage and recovery risks.
The gap is especially important in Entra ID. Microsoft’s own documentation notes that Conditional Access for workload identities applies to service principals but does not cover managed identities. That does not mean managed identities cannot be contained, but it does mean the available controls, logging and break-glass procedures differ. A security product that claims universal agent containment must make those differences visible rather than collapsing all machine identities into one category.
Before enabling automated action, teams should require a documented answer to four practical questions:
- The product should identify the exact directory object, permissions and authentication method it will act on before any automated isolation runs.
- The containment playbook should distinguish a service principal from a managed identity, an application registration and a human user account.
- Every containment action should be logged to the organization’s normal incident record, including Microsoft Sentinel or another central SIEM if that is part of the SOC workflow.
- The organization should test how it restores legitimate automation after isolation, because shutting down an agent identity can also halt deployments, workflows, integrations or line-of-business processes.
Quest says its platform works alongside Microsoft Sentinel, Security Copilot and Defender for Identity. That integration language does not, by itself, show which actions can be initiated from which console, whether telemetry is bidirectional, or whether a customer must license separate Quest modules. Those are procurement and design questions, not implementation trivia.
Secure Replay is the new feature with the biggest recovery promise — and the least public evidence
Secure Replay is the one feature Quest describes as being in private preview. The company says it uses AI to distinguish malicious changes from legitimate activity and aims to return an organization to the last known safe state instead of restoring an older backup wholesale.
For a compromised Active Directory or Entra ID tenant, that is an ambitious goal. Recovering identity systems is difficult precisely because a directory is not a static data store. An incident may involve changes to group membership, privileged role assignments, application consent, credentials, Conditional Access policies, federation settings, device records, authentication methods or service-principal permissions. Restoring everything to a previous point can erase legitimate business changes; restoring too selectively can retain attacker persistence.
The concept behind Secure Replay is sensible: preserve normal operations while rolling back changes identified as malicious. But Quest has supplied no public technical description of its detection model, supported change types, confidence thresholds, approval workflow, forensic evidence requirements or failure modes. Nor has it published a test case showing a replay of a real Entra or Active Directory compromise.
That makes Secure Replay a preview capability to evaluate, not a recovery assurance to assume. A product that decides which directory changes are malicious must be held to a high standard, particularly when it may influence access control for administrators and business applications. The first proof point should be whether a customer can simulate credential theft and privilege escalation in a nonproduction tenant, inspect the proposed rollback set, reject incorrect recommendations, and produce an auditable record of every restoration decision.
Quest’s assertion that its technology can improve recovery time by up to 90% compared with enterprise backup tools is also the company’s own research claim. Neither Security Today nor Portal ERP provides methodology, baseline environments, customer sample size or independently audited measurements. It should not be read as a general benchmark for Entra or AD recovery planning.
Guardian services acknowledge the harder problem: recovery is organizational
Quest also introduced Guardian Managed Recovery Services for Active Directory and Entra ID, with a partner-delivered, white-label option. This is less glamorous than autonomous response, but it addresses a persistent failure point: an organization can own backup software and still fail to restore identity services coherently during a major incident.
Quest’s own 2026 identity threat detection and response survey found that more than 75% of respondents lacked a tested recovery plan. Because that is vendor-sponsored research, it cannot establish a market-wide rate on its own. Still, the operational lesson is sound: identity recovery needs rehearsed ownership, known restoration order, clean administration paths and tested communications with the teams whose applications depend on authentication.
A recovery service can provide expertise, but it does not eliminate the customer’s responsibility to define an authoritative recovery plan. Windows and Entra administrators should know which domain controllers, privileged accounts, break-glass identities, identity connectors, application registrations and federation dependencies are in scope. They should also know whether their recovery procedure assumes that Microsoft 365, Azure resources, on-premises AD and hybrid synchronization remain available at the same time.
Quest’s Secure Migration addition fits the same theme. The company says it can flag excessive privileges, inactive accounts and vulnerable devices before a migration or consolidation. That may help reduce inherited identity debt during mergers, AD modernization and tenant changes, but it is preventive hygiene rather than evidence that the new agent-defense controls can stop a live compromise.
What Quest has announced — and what customers still need to verify
Quest’s expansion is a credible indicator that AI-agent governance is converging with the less fashionable work of service-account management, privileged-access review and directory recovery. The company’s Microsoft focus is real: its platform was introduced in April 2026 around identity threat detection and response, secure identity modernization, AD and Entra recovery, and it later received a FedRAMP High authorized offering for certain identity defense and recovery capabilities.
But the September announcement is best treated as a product roadmap and evaluation event, not proof that a new category of AI risk has been solved. The reports from Security Today and Portal ERP repeat Quest’s feature descriptions and claims; they do not name a customer deployment or independently validate the autonomous isolation and AI-assisted recovery functions.
For organizations already running Microsoft Entra workloads or deploying agents against Microsoft 365, Azure and internal applications, the immediate task is more concrete than buying a new “agent security” product: inventory the service principals, managed identities, app registrations, delegated permissions and credentials that agents already use. Then test whether existing controls can revoke access quickly, preserve forensic logs and restore a known-good identity state. Quest’s new platform may become part of that process, but Secure Replay’s private-preview status and the absence of published availability and licensing detail mean it is not yet a substitute for a tested Entra and Active Directory recovery plan.