How can organisations reduce lateral movement without breaking normal operations?

Why This Matters for Security Teams

Reducing lateral movement without disrupting operations is a core NHI and agentic control problem: identities need enough reach to complete work, but not enough freedom to roam. Static allowlists and coarse network segmentation often fail because service accounts, API keys, and AI agents do not behave like users. They can chain tools, retry failed actions, and pivot across systems in ways that are invisible until an incident occurs.

That is why current guidance increasingly favours identity-centric containment rather than blanket blocking. The MITRE ATT&CK Enterprise Matrix helps teams map how lateral movement actually unfolds in enterprise environments, while NHIMG’s Ultimate Guide to Non-Human Identities notes that 97% of NHIs carry excessive privileges, which directly widens the blast radius when one identity is misused.

In practice, many security teams encounter lateral movement only after an overprivileged identity has already been reused across systems, rather than through intentional testing of access boundaries.

How It Works in Practice

The most effective approach combines behavioural learning, least privilege, and just-in-time exception handling. Start by establishing what “normal” looks like for each identity class: which hosts it should reach, which APIs it should call, which data stores it should touch, and at what frequency. For NHI-heavy environments, that profile should be tied to workload identity, not just a credential string. Standards such as MITRE ATT&CK Enterprise Matrix help teams model common movement paths, while NHIMG’s 52 NHI Breaches Analysis shows how compromise often expands when one identity can move laterally without being noticed.

Operationally, teams usually combine these controls:

• Granular identity policies that scope access by service, environment, and task, not broad role membership.
• JIT credentials with short TTLs so access expires automatically after the task completes.
• Context-aware policy evaluation at request time, including source workload, destination, time, and requested action.
• Exception workflows for legitimate break-glass activity, with explicit approval and time-bounded access.
• Continuous telemetry to compare actual behaviour against the learned baseline and flag drift.

For agentic systems, this is especially important because an AI agent may discover a new tool path or retry sequence that a static policy never anticipated. Guidance from Storm-2949 Azure Breach reinforces how a single identity can become a broad compromise path once trust boundaries are too loose.

These controls tend to break down when legacy applications depend on shared service accounts or when network location, not identity, is still the primary trust signal, because the policy engine cannot distinguish legitimate reuse from lateral abuse.

Common Variations and Edge Cases

Tighter containment often increases operational overhead, so organisations have to balance blast-radius reduction against change friction and support load. That tradeoff is real in CI/CD pipelines, batch jobs, and multi-service workflows where identities legitimately touch many systems within a short window. Best practice is evolving, but there is no universal standard for this yet: some environments can enforce per-request authorisation cleanly, while others need staged rollouts and exception registers before they can get there.

One common edge case is shared infrastructure tooling. If a deployment pipeline or orchestration agent uses the same identity across environments, overly narrow policies can break releases. In those cases, use environment-scoped identities, stronger secret rotation, and explicit destination constraints rather than widening access across the board. Another edge case is third-party integration, where external systems may need temporary access to internal APIs. NHIMG’s research also shows that 92% of organisations expose NHIs to third parties, so partner access should be treated as a separate trust tier, not folded into internal policy.

For agentic AI and autonomous workflows, the safest pattern is to permit only the minimum tool chain needed for the current task, then revoke immediately after completion. That keeps normal operations moving while reducing the chance that one compromised identity can pivot into adjacent systems.

Related resources from NHI Mgmt Group

• How should healthcare organisations reduce VPN overreach without slowing clinical access?
• How can organisations reduce secret leakage in ServiceNow at scale?
• How do organisations reduce false positives in secret detection pipelines?
• How can organisations reduce over-privileged OAuth access without breaking business workflows?