AI-fueled OT exposure is widening risk for fuel infrastructure

At a glance

What this is: This is Armis's analysis of suspected Iranian attacks on U.S. gas-station Automatic Tank Gauges, where exposed OT devices were probed, display readings were tampered with, and deeper access was attempted.

Why it matters: It matters because OT assets can become direct operational and safety liabilities when internet exposure, weak identity controls, and poor visibility leave critical infrastructure open to reconnaissance and abuse.

Context

Operational technology becomes a security problem when systems built for reliability are left directly reachable from the public internet. In that condition, attackers do not need sophisticated access paths; they can find exposed devices, test default trust assumptions, and probe for control over physical processes.

The article centers on a wider governance gap in critical infrastructure: security teams cannot protect what they have not inventoried, segmented, or placed behind explicit access controls. That problem is broader than fuel stations, because any exposed OT asset can become a route from cyber reconnaissance to operational disruption.

Key questions

Q: What breaks when OT devices are exposed to the internet?

A: Internet exposure turns HMIs and PLCs into reachable control points rather than protected field assets. Once attackers can authenticate or exploit them, they can alter process settings, change credentials, and disrupt monitoring. The practical failure is not only compromise of a device, but loss of trust in the control plane that keeps physical operations stable.

Q: Why do weak credentials and flat networks make OT attacks worse?

A: Weak credentials and flat networks turn a limited compromise into a wider operational risk. If a device trusts default passwords or sits on the same network as business systems, attackers can move from discovery to control with far fewer barriers. In OT, that can translate into manipulated readings, disabled alarms, or interference with physical processes. Segmentation and identity controls matter because they prevent one exposed device from becoming a route to the whole environment.

Q: How can security teams tell whether an OT assessment is actually working?

A: An assessment is working when it produces approved, executable remediation plans rather than static reports. Look for evidence that findings are being re-ranked using production impact, exploitability, and compensating controls, and that operations teams are accepting the resulting actions. If the output is still a spreadsheet, the process is not yet effective.

Q: What should organisations do when legacy OT must remain remotely accessible?

A: They should force remote access through a controlled gateway, require unique credentials, and isolate the OT segment from corporate and payment systems. The decision should be made as a governance exception, not as an informal convenience. If remote access is unavoidable, the access path must be explicit, monitored, and limited to the smallest practical scope. Anything less leaves operational systems too easy to reach and too hard to defend.

Technical breakdown

Why exposed ATGs create an internet-facing control problem

Automatic Tank Gauges are not just level indicators. They also monitor leaks, coordinate inventory data, and can influence operational decisions tied to fuel handling and alarms. When these devices sit on the public internet, the attack surface shifts from a protected control environment to an addressable service that anyone can enumerate. Read-only exposure reduces the attacker’s options, but it does not remove reconnaissance value, because it still reveals device behaviour, firmware patterns, and potential management interfaces. The deeper problem is that legacy OT often lacks the assumptions modern security tools rely on, such as strong authentication, continuous logging, and well-defined ownership.

Practical implication: Treat exposed OT telemetry and management surfaces as security-relevant assets, not benign instrumentation.

How reconnaissance becomes operational pressure in OT environments

OT reconnaissance usually starts with broad scanning, then moves to device fingerprinting, access testing, and protocol-specific probing. In this article, the reported use of Shodan and Censys shows the value of discovery at internet scale: once devices are exposed, they can be catalogued, sampled, and tested repeatedly. AI changes the economics by accelerating that triage, helping attackers sort weak devices from harder targets and generate exploitation ideas faster than manual analysis would allow. Even when an attacker cannot fully compromise the device, reconnaissance can still shape later targeting by revealing where trust is misplaced and where isolation is absent.

Practical implication: Build detection and inventory processes that assume hostile scanning will happen before exploitation does.

Why legacy OT fails when identity and segmentation are weak

The core failure is architectural, not merely technical. Legacy industrial systems were designed for uptime and convenience, so they often inherit default credentials, flat network paths, and minimal separation between business and operational zones. Once a payment or administrative environment can reach an OT segment, the boundary between cyber compromise and physical consequence erodes. In this kind of environment, network segmentation and identity controls are not separate disciplines. They are the controls that stop a compromise in one domain from becoming a problem in another, especially when remote polling or vendor access is still required.

Practical implication: Enforce segmentation and unique credentials before expanding any remote access path into OT.

Threat narrative

Attacker objective: The objective is to gain enough access to interfere with fuel infrastructure, disrupt monitoring, and create operational or economic damage without needing an internal compromise.

1. Entry occurred through internet-facing ATGs that were discoverable via public scanning and could be reached without an internal foothold.
2. Credential and access probing followed as the attackers tested the devices, interacted with deception systems, and attempted to learn how to operate them with AI assistance.
3. Impact was limited in the reported cases, but the same exposure pattern could enable alarm tampering, leak-monitoring disruption, or broader operational sabotage.

Breaches seen in the wild

• Colonial Pipeline ransomware attack: A dormant VPN account with a leaked password and no MFA let DarkSide shut down a major US fuel pipeline for six days.

Read and download The State of NHI & AI Agent Breach Report 2026, covering 200+ breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Complete asset awareness is now a foundational control for OT security. The article’s central point is not that attackers found a novel exploit, but that exposed assets remained visible and reachable in the first place. In critical infrastructure, unmanaged internet exposure turns inventory failure into operational risk. That makes discovery, ownership, and segmentation baseline governance requirements, not optional hardening steps. Practitioners should treat unknown OT assets as a material exposure until they are identified, classified, and isolated.

AI is reducing the cost of OT reconnaissance faster than defenders are reducing exposure. The attackers did not need to break a hardened perimeter if the devices were already publicly discoverable. That changes the economics of defense: the gap is no longer just detection speed, but whether a device was ever meant to be internet-facing at all. For security programmes, this is a visibility and exposure management problem as much as a malware problem. Practitioners should assume scalable reconnaissance will keep compressing response windows.

The identity gap in OT is often the point of failure, not the afterthought. The article’s recommendation to remove factory-default passwords and place remote access behind a secure gateway is a reminder that legacy devices still depend on basic identity hygiene. Where OT systems lack unique credentials, strong authentication, or clear access ownership, the path from discovery to control becomes far shorter. That makes credential policy and segmentation inseparable in industrial environments. Practitioners should review OT access as a governed identity boundary, not a convenience layer.

Exposed legacy control systems create a governance debt that defenders keep paying at runtime. The article describes a classic condition where systems were built for availability and later forced into adversarial environments without a corresponding control redesign. That debt shows up as flat networks, weak authentication, and incomplete monitoring. The practitioner implication is straightforward: if the asset model is not accurate, the control model will fail under pressure.

What this signals

Exposure management is becoming the deciding control for OT resilience. The article shows that the first problem is not exploitation, but discoverability. If a critical asset can be found instantly on the open internet, every later control is trying to recover lost ground. Practitioners should expect the boundary between asset management and security operations to keep narrowing as attackers automate reconnaissance.

Legacy OT now needs identity governance as much as network filtering. Default passwords, unmanaged remote access, and unclear ownership are the conditions that let exposure become compromise. That means industrial environments need explicit access boundaries, not just perimeter devices. When the identity layer is weak, segmentation becomes harder to trust and incident scope becomes harder to contain.

Asset visibility is the new minimum viable control for critical infrastructure. Organisations cannot protect what they cannot continuously see. In practice, that means every new device, remote access path, and unexpected internet exposure must be treated as a governance event, not a discovery surprise.

For practitioners

• Map every internet-facing OT asset Create and maintain a live inventory of ATGs and adjacent industrial devices, including ownership, exposure state, and business criticality.
• Remove public internet exposure Place any required remote polling or administration behind a secure VPN gateway and block direct public access to OT devices.
• Eliminate shared or factory credentials Assign unique, complex passwords to every device and prohibit default credentials on any operational technology system.
• Segment OT from corporate and payment networks Separate back-office systems from operational networks so a compromise in one zone cannot directly reach fuel infrastructure.
• Monitor for hostile scanning and unexpected device changes Use continuous asset discovery and contextual risk assessment to detect exposed devices, new internet routes, and abnormal device behaviour.

Key takeaways

• The article’s main risk is not a single exploit, but the persistent exposure of operational technology that was never meant to be freely discoverable on the internet.
• Reported attacker activity included probing and display tampering, which illustrates how visibility into exposed devices can be enough to create operational risk even before full compromise.
• The limiting control is accurate asset inventory tied to segmentation and identity controls, because unmanaged reachability is what turns OT into a target.

Related NHI Mgmt Group resources: AI Infrastructure Workload Identity Guide · Threat Intelligence Guide · Colonial Pipeline ransomware attack

Key terms

• Operational Technology: Operational Technology is the hardware and software that monitors or controls physical processes such as manufacturing lines, utilities, and transportation systems. Unlike standard IT, OT prioritises uptime and safety, so identity controls must be precise enough to reduce risk without interrupting essential operations.
• Asset Exposure: Asset exposure is the extent to which a vulnerable system is reachable by an attacker in practice. It includes internet reachability, routing mistakes, shadow infrastructure, and other conditions that turn a theoretical flaw into an actual attack path. Exposure is often the deciding factor in prioritization.
• Network Segmentation: Network segmentation divides traffic and resources into controlled zones so access can be restricted between groups, systems, or applications. In remote access design, segmentation limits what a connected user or workload can reach after authentication, which reduces lateral movement and shrinks blast radius.
• Exposure management: Exposure management is the practice of identifying which assets are reachable by attackers and reducing that reach before exploitation occurs. For collaboration systems like SharePoint, it is not enough to know that a patch exists, because public accessibility changes the speed and likelihood of attack.

Deepen your knowledge

NHI Mgmt Group covers identity security, NHI governance, and agentic AI through independent research, practitioner guides, and the NHI Foundation Level course. Explore nhimg.org for resources that connect identity governance to the broader security disciplines your programme depends on.