The costliest cyber incidents in pharmaceutical manufacturing this year did not begin on the plant floor. They began in IT, and production stopped anyway. For CISOs responsible for GxP systems, that is the core problem: validated, hard-to-patch OT environments sit behind enterprise networks that attackers now treat as the front door. 

Why GxP production is a different OT problem 

OT security advice written for power plants or water utilities breaks down in a pharmaceutical plant, and the reason is validation. A GxP-regulated production system is qualified in a specific, documented state. Change that state, whether by installing an endpoint agent, applying a patch or re-addressing a controller, and you may owe the quality unit a change control, a risk assessment and a revalidation. One OT security vendor puts the cost of revalidating a single system between $50,000 and $500,000, a figure worth treating as a vendor estimate but one that matches what plant teams describe. 

So the usual playbook, patch fast and put an agent on every host, collides with the rules that keep medicines safe. That is why many pharma sites still run unsupported operating systems on historians, HMIs and lab instruments. It is not negligence. It is a trade-off between two compliance obligations. 

There is a second difference: what an attacker can do to you. In most industries the damage is downtime. In pharma, the damage can include questions about the integrity of every batch record produced while a system was compromised. Regulators expect data to be attributable, complete and protected from deletion, and an intruder with admin rights breaks all three at once. A cyber incident can become a compliance crisis the same afternoon. 

Three consequences follow for how a security program should be designed: 

• Protection has to be network-first. Controls that sit around a validated system, such as segmentation, brokered access and passive monitoring, usually avoid touching the qualified state. 

• Security and quality must share a change process. A firewall rule on a GxP network is a GxP change. Treat it that way from the start. 

• Recovery means restoring a validated state, not just a running one. A restored server that cannot prove its configuration matches the qualified baseline is not ready for production. 

What changed recently: incidents and threat data 

Three incidents at life sciences manufacturers this year show how disruption now arrives, and none of them required an attacker to reach a PLC. 

The downstream effects reached patients. After the Stryker attack, NHS England asked its partners to prioritize clinically important supplies and assess their dependence on the company's products. Supply-chain dependence is now a clinical risk, not only a commercial one. 

The wider numbers 

The ICS Cybersecurity Report, covering 2025, describes adversaries moving from reconnaissance toward mapping control loops, the logic that governs how a physical process is run. Independent summaries of the same dataset report: 

• Ransomware hit roughly 3,300 industrial organizations in 2025, driven by 119 groups, up from 80 the year before. Manufacturing made up more than two-thirds of victims. 

• In Dragos field assessments, 81% found poor IT/OT segmentation, and 73% of its incident response cases over the years involved compromised VPN or jump-host credentials. 

• Quarterly tallies add up to about 2,160 industrial ransomware incidents in the first half of this year, around 58% more than the same period a year earlier. That last comparison is Stingrai's own calculation from Dragos quarterly data, and it counts victims named on leak sites, not every attack. 

Treat those as directional. The pattern, though, is consistent across sources: the entry points are ordinary (stolen credentials, remote access, flat networks), and the physical consequences are not. 

Regulatory pressure: what is in force and what is coming 

The single most useful thing to tell your quality and compliance colleagues is that security is moving inside GMP. The table separates what binds you today from what is still moving. 

The practical point: do not wait for the final Annex 11 before acting. The draft's direction, security as a lifecycle requirement tied to identity, audit trails and supplier oversight, matches what inspectors already ask for under the 2011 text and what FDA letters already cite. Companies that build to the draft now will not need a second program later. 

Where GxP plants are actually exposed 

Walk enough pharma sites and the same weaknesses repeat. None are exotic. 

A flat or loosely segmented IT/OT boundary. MES, electronic batch records, SCADA and historians usually depend on shared services such as Active Directory, DNS and patch infrastructure. Compromise those from the IT side and you can lose the plant without touching a controller. The West Pharmaceutical case, where the intrusion reportedly stayed in IT but production stopped, is the textbook version. 

Remote access that outlived the project. Equipment vendors and integrators need access for support, and that access tends to persist as always-on VPN accounts or jump hosts with shared credentials. Dragos reports that compromised VPN or jump-host credentials appear in most of its incident response cases, as noted above. 

Removable media and engineering laptops. Validated systems often cannot run endpoint agents, so USB drives and vendor laptops become the practical way to move data and files. TXOne cites removable media as the trigger for about 25% of top OT security incidents it analyzed, though as a vendor figure it should be read as indicative. 

Legacy laboratory and process equipment. Chromatography stations, balances, filling-line controllers and quality lab instruments frequently run operating systems that no longer receive vendor support. Replacing them means revalidation and often new capital approval, so they stay. 

Identity weaknesses that double as data integrity findings. Shared accounts on lab instruments, analysts with delete rights and unrecorded password handling appear again and again in FDA warning letters. Fixing them is both a security improvement and a compliance fix, which makes it the easiest budget conversation to win. One FDA letter to an Indian API maker described passwords handwritten in an uncontrolled notebook. 

Contract manufacturers and extended supply chains. As more production moves to CMOs and CDMOs, your exposure includes sites you do not run. Rockwell Automation frames the question well: how do you keep a flexible supply chain and manage cyber risk when manufacturing extends beyond your walls? NIS2 now makes supplier assurance an explicit duty for in-scope entities. 

Sector-level data points the same way. TXOne describes pharma as the third most targeted sector in Europe for OT threats, and Seqrite's India Cyber Threat Report found healthcare and pharma among India's most targeted sectors in 2025. Both are vendor-published reports, so check methodology before quoting figures in board material. 

A control playbook that respects validation 

The SANS five ICS critical controls are a sound backbone for pharma and chemical plants. The sequence below adapts them to environments where every change needs a quality sign-off. 

1. Build the asset inventory around GxP impact, not just IP addresses. For each asset record the system owner, validation status, GAMP category, OS and support status, and which batch records or critical quality attributes it touches. This is the document that lets security and quality talk about the same risk. 

2. Segment first, because segmentation rarely changes a validated system. Separate enterprise IT, the site DMZ, the MES/batch layer and the control layer, and allow only the flows each function needs. Treat firewall rules as change-controlled GxP records. 

3. Replace standing remote access with brokered, time-boxed sessions. Vendor access should require named accounts, MFA, approval for each session and a recording. Remove shared credentials. This addresses the most common entry path in Dragos incident data. 

4. Use compensating controls for what cannot be patched, and document them. IEC 62443-2-3 expects deferred patches to carry documented compensating controls and periodic reassessment. For end-of-life or sealed systems that means identity-based segments, allow-listed flows, brokered access and virtual patching, with the control dated and reviewable at audit. 

5. Monitor passively. Network-based monitoring that mirrors traffic watches PLCs, historians and engineering workstations without installing anything on them. Alert on new devices, new protocol commands and changes to controller logic. 

6. Fix identity and audit trails together. Eliminate shared accounts on lab and process systems, enforce least privilege, remove delete rights, and make sure audit trails are protected from the people they record. One project, two sets of findings closed. 

7. Control removable media and engineering laptops. Use scanning stations or kiosks at site entry points, keep vendor laptops in a managed pool, and log every transfer into validated networks. 

8. Protect the identity infrastructure your plant depends on. Harden the domain controllers and cloud endpoint management that MES, SCADA and laptops rely on. The Stryker incident showed that control of a management plane can mean destruction at scale. 

9. Back up for validated recovery. Keep offline, tested backups of controller logic, HMI projects, recipes, batch-system databases and golden images, plus the evidence that a restored system matches its qualified baseline. Version control and automated disaster recovery for OT assets are the core elements of several pharma-focused OT tools. 

10. Extend assurance to CMOs, CDMOs and OT vendors. Add security clauses to quality agreements, ask for segmentation and remote access evidence during audits, and track supplier incidents the way you track supplier deviations. 

An eleventh item belongs on the board's agenda: decide in advance which systems you would deliberately isolate, and who is authorized to do it. West Pharmaceutical's global shutdown was precautionary. Plants that rehearse that call recover faster than those making it for the first time at 2 a.m. 

Incident response when batches are at stake 

A generic OT incident plan leaves out the questions the quality unit will ask within the first hour. Build these into the plan and rehearse them in a tabletop. 

• Open a GMP deviation alongside the incident ticket. The cyber response team and the quality unit should work from the same timeline from the start, because batch disposition decisions depend on it. 

• Define the isolation decision tree. For each line or area, who can disconnect it from the network, what happens to in-process batches, and what safe state the equipment reaches. Precautionary shutdowns are common, so the decision needs an owner. 

• Preserve evidence without breaking data integrity. Capture logs and images in a way that keeps audit trails intact and defensible to inspectors. 

• Assess data integrity impact explicitly. Which records could have been altered, deleted or made unavailable? Which batches were in process? The answers drive release decisions, not the IT recovery schedule. 

• Restore to a qualified state. Rebuild from known-good images and verified configurations, then document that the system matches its validated baseline before returning it to production. Involve validation early so recovery is not delayed by paperwork discovered late. 

• Plan disclosure paths in parallel. Depending on your footprint, that may include NIS2 reporting, securities filings and regulators. West Pharmaceutical and Stryker both disclosed through regulatory filings within days, and NIS2 reporting runs alongside, not instead of, pharmacovigilance, GDP and GMP deviation obligations. 

Run at least one tabletop a year that includes the plant manager, head of quality, a validation lead, legal and an executive sponsor. A scenario worth testing: the plant's IT estate is encrypted, controllers are untouched, and the site must decide in two hours whether to keep running. 

Frequently asked questions 

What is OT security in pharmaceutical manufacturing? 

It is the protection of the systems that run and record drug production: PLCs, DCS and SCADA, MES and electronic batch records, historians, and laboratory instruments. The difference from other industries is that these systems are validated, so security controls must be introduced through change control without disturbing their qualified state. 

Does GxP validation prevent patching? 

It makes patching slower and more expensive, but it does not forbid it. Patches go through risk assessment, testing and change control. Where that is not feasible, IEC 62443-2-3 accepts documented compensating controls, such as segmentation and monitoring, in place of the patch. 

Is the revised EU GMP Annex 11 final? 

Not yet. The 2011 version remains the one in force. The revised draft, which puts security and identity management at the center, was published in July 2025, and analysts expect finalization between late 2026 and 2027. Check the European Commission's EudraLex Volume 4 page for the current status. 

Does NIS2 apply to pharmaceutical manufacturers? 

In most cases, yes. Manufacturers of basic pharmaceutical products and pharmaceutical preparations are listed in the health sector of Annex I. Large entities are generally essential entities and medium-sized ones important entities, but classification and registration depend on national law, so confirm with your member state's competent authority. 

What are the most common entry points into pharma OT? 

Compromised remote access credentials, flat networks that let IT incidents spread, removable media, and unmanaged legacy hosts. Dragos reports that poor IT/OT segmentation was found in a large majority of its assessments and that compromised VPN or jump-host credentials feature in most of its response cases. 

Which standards should a pharma OT security program follow? 

Most programs combine ISA/IEC 62443 for OT security architecture and patch management, NIST SP 800-82 for ICS guidance, the SANS five ICS critical controls as a prioritization lens, and GAMP 5 for the validation side. Regulatory drivers are FDA CGMP, EU GMP Annex 11 and, in the EU, NIS2. 

Can a cyber incident trigger a batch recall or regulatory action? 

It can affect batch release if data integrity cannot be demonstrated. Whether a recall or regulatory notification follows depends on the facts, so involve the quality unit and regulatory affairs from the first hours. 

How Shieldworkz supports pharma OT security 

Shieldworkz is an end-to-end industrial OT cybersecurity company, covering SCADA, OT and IoT environments across critical infrastructure with products, services and solutions delivered as one stack. For pharmaceutical manufacturers, that matters because the hardest problems sit between the layers: asset visibility that feeds change control, patch decisions that need compensating controls, and recovery that has to land on a validated baseline. 

• Visibility and monitoring: the OThello platform gives security and plant teams a shared view of OT assets and activity, supporting the inventory and passive monitoring steps above. 

• Removable media risk: MediaScan addresses the USB and engineering-laptop path into validated networks. 

• Patch decisions in OT: OT-specific patch management helps teams decide what to patch, what to defer and what compensating controls to document. 

• Services: assessments, architecture and segmentation design, and incident readiness exercises that include plant, quality and validation stakeholders. 

To see where your GxP sites stand against the playbook above, talk to the Shieldworkz team about an OT security assessment. 

Additional resources:

Incident Response Plan for Municipal & Wastewater Utilities here
What Is Removable Media? Risks, Policies, and Industrial OT Security Solutions here
Free Removable Media Policy Template for OT and IT Teams here
Remediation Guides here