Ransomware Hit the Delivery Mechanism for 70% of the World's Injectable Drugs

Ransomware Hit the Delivery Mechanism for 70% of the World’s Injectable Drugs

Plus: EU denies medical devices the AI Act exemption industrial AI won, Roche acquires your pathology vendor, and FDA’s HALO rewards structured submissions

Week of May 11–17, 2026  ·  ~13 min read  ·  Compiled with Perplexity and Claude AI.

Three converging shocks this week:

• West Pharmaceutical Services confirmed a ransomware attack encrypting systems globally and exfiltrating data at an S&P 500 supplier whose injectable packaging components reach approximately 70% of the world’s injectable drugs — the second major pharma-adjacent manufacturer compromised in H1 2026
• The EU Digital Omnibus reached a provisional agreement May 7, extending AI Act high-risk deadlines — but medical device manufacturers did not receive the dual-compliance exemption that industrial AI secured, leaving them facing parallel obligations under both the AI Act and MDR/IVDR
• Roche’s $1.05B acquisition of PathAI is removing one of the last independent computational pathology platforms from biopharma’s vendor landscape, forcing organizations using PathAI for clinical trial pathology to reassess data governance and vendor independence

The connecting thread: infrastructure choices made in the last 18 months — vendor dependencies, data architecture, documentation quality, cyber segmentation — are being stress-tested simultaneously.

🤖 AI & Data

This week’s AI deals: major pharma committing to multi-year, production-scale partnerships while independent AI platforms continue consolidating under large diagnostic and biopharma acquirers.

Roche Acquires PathAI for Up to $1.05B — Vendor Independence in AI Pathology Is Now Scarce

On May 7, Roche announced a definitive agreement to acquire PathAI for $750M upfront and up to $300M in milestone payments, integrating the acquired entity into Roche Diagnostics and targeting H2 2026 close.

What happened:

• PathAI’s AISight IMS supports AI-powered workflows for digital pathology, biomarker discovery, patient stratification, and companion diagnostic development — embedded in active clinical programs at multiple large biopharma companies
• Roche is a direct competitor in companion diagnostics and oncology therapeutics, creating an immediate conflict-of-interest question for biopharma organizations using PathAI services for non-Roche sponsored trials

Why it matters to you:

• Contract terms, data governance provisions, and IP protection for pathology data flowing through AISight must be reviewed before the deal closes — independent computational pathology vendors are increasingly scarce
• Oncology-focused biotechs and CDMOs with PathAI-dependent clinical trial programs face a vendor dependency risk requiring an immediate alternative-vendor assessment

AstraZeneca Expands Immunai Collaboration to $37.5M — Three Renewals Signal Production-Grade Clinical AI Is Viable

On May 7, Immunai announced a third expansion of its oncology AI collaboration with AstraZeneca, eligible for up to $37.5M through 2026–2027, using the AMICA-OS platform across oncology programs for biomarker discovery, patient stratification, mechanism-of-action analysis, and dose optimization.

What happened:

• Three multi-cycle renewals signal production-grade AI for clinical immunology is operationally viable in regulated environments — AMICA-OS integrates with CDM systems, genomics repositories, single-cell sequencing pipelines, and regulatory-grade analytical environments
• Sponsors increasingly expect CDMO and CRO partners to process and return AI immunology analysis in compliant, auditable formats — understanding AMICA-OS-class pipeline architecture is becoming a contract capability requirement

Why it matters to you:

• AMICA-OS combines proprietary data assets with foundation AI model infrastructure — a combination difficult to replicate quickly in-house and that generates a strategic dependency worth tracking in your vendor landscape
• Informatics infrastructure unable to interface with single-cell immune profiling datasets and foundation model outputs will be a constraint in partnership negotiations within 12–18 months

Recursion Q1 2026: $665M Cash Runway and the Clearest Case Study in AI-Native Drug Discovery at Scale

Recursion Pharmaceuticals reported Q1 2026: $6.5M revenue (down from $14.7M due to milestone timing), net loss narrowed to $117.5M (versus $202.5M), and $665.2M cash supporting runway into early 2028.

What happened:

• REC-4881 showed 43% median polyp-burden reduction at Week 13, deepening to 53% at Week 25; FDA engagement initiated for a potential registrational pathway with an update expected H2 2026; REC-4539 dosed its first Phase 1 patient
• CEO Najat Khan disclosed an in-house AI model screening 300 million patients to identify optimal trial sites; partnered programs with Sanofi and Genentech/Roche now exceed $500M in cumulative upfront and milestone payments

Why it matters to you:

• Recursion is the clearest current case study of a vertically integrated AI-native drug discovery platform at scale: biological imaging data, proprietary AI models, in-house supercomputing with NVIDIA, and automated laboratory infrastructure in a single stack
• The post-Exscientia integration is a reference case for data, compute, and governance challenges any CIO will face in AI biotech M&A — directly relevant to build-vs-buy-vs-partner decisions

⚖️ Regulatory & Policy

Three developments this week share a common implication: the quality of your documentation architecture is now a direct input to regulatory speed and enforcement exposure.

EU Digital Omnibus Confirms Dual AI Act + MDR/IVDR Compliance for Medical Devices — The Exemption Industrial AI Won, Medtech Did Not

EU co-legislators reached a provisional Digital Omnibus agreement May 7, extending AI Act high-risk compliance deadlines: standalone Annex III systems move from August 2, 2026 to December 2, 2027; AI embedded in regulated products including medical devices and IVDs moves to August 2, 2028. Industrial AI secured a dual-compliance exemption; medical technologies did not, despite lobbying by MedTech Europe.

What happened:

• Medical technologies remain subject to parallel compliance under both the AI Act and MDR/IVDR — requiring separate conformity assessments, different documentation formats, and different audit logic for the same AI system
• The agreement narrows the “safety component” definition (AI that only assists users without creating safety risks may avoid high-risk obligations), but this requires case-by-case legal analysis; Article 50 transparency obligations remain on schedule for August 2, 2026

Why it matters to you:

• Organizations that planned to rely on a single MDR-based pathway must now budget for parallel AI Act documentation alongside MDR/IVDR Technical Documentation requirements
• The provisional agreement is not yet formally adopted — the August 2026 deadline remains in law until formal adoption before June 2026; compliance programs cannot pause on an unadopted deal

FDA’s First GMP AI Warning Letter Sets the Human Oversight Standard for Regulated Document Workflows

An April 2 FDA Warning Letter (WL 320-26-58) to Purolea Cosmetics Lab, analyzed by EAS Consulting and Clarkston Consulting, marked the first FDA citation for “inappropriate use of artificial intelligence in pharmaceutical manufacturing” as a standalone GMP compliance observation under 21 CFR 211.22(c).

What happened:

• The facility used AI agents to generate drug product specifications, procedures, and master production records without qualified human review; when investigators found no process validation had been conducted, the company’s response that the AI “never told it” the requirement applied was found unacceptable
• FDA’s stated position: AI may aid document creation, but the Quality Unit’s approval authority under 21 CFR 211.22 is non-delegable to an AI system regardless of competence; Computer Software Assurance (CSA) validation of AI document generation tools is explicitly required

Why it matters to you:

• This is FDA’s first articulation of enforcement position on AI in regulated manufacturing workflows — it applies equally to large pharma, biotech, CDMO, and CRO environments deploying generative AI in quality systems
• The Clarkston compliance checklist: GxP risk classification → hallucination check → independent SME review → QA/RA review → ALCOA+ data integrity → inspectability of generation process → CSA validation

EUDAMED Mandatory Registration Arrives May 28; FDA’s AI Inspection Pilot Has Completed 46 Assessments

Two converging regulatory milestones this week. EUDAMED becomes mandatory for new medical devices on May 28, 2026, integrating device registration, UDI, vigilance reporting, and post-market surveillance into a single EU platform; legacy devices have until November 27, 2026. FDA’s AI-informed one-day inspection pilot has completed approximately 46 assessments using risk stratification to identify lower-risk facilities for abbreviated review.

What happened:

• EUDAMED registration accuracy becomes an ongoing compliance function — not a one-time event — as regulators across EU member states access a unified, live device database
• FDA has flagged discrepancies between registered and actual operations as a key AI risk signal — a direct concern for organizations that have undergone M&A, site transitions, or product changes since their last inspection

Why it matters to you:

• Both platforms create pressure for real-time documentation accuracy: EUDAMED requires current registration data and FDA’s AI risk scoring means documentation gaps are actively influencing inspection allocation
• The combination of AI-informed site selection and HALO’s submission cross-referencing capability means inconsistencies between registered information and actual submission history are no longer theoretical — they are triggering regulatory attention

🔒 Cybersecurity & Risk

West Pharmaceutical’s ransomware attack crystallizes a 2026 pattern: pharma-adjacent manufacturers with high operational interdependence to their customers are priority ransomware targets, and enterprise IT disruption translates directly into production shutdown.

West Pharmaceutical Services Ransomware Attack Disrupts Global Injectable Drug Supply Chain

On May 4, West Pharmaceutical Services (NYSE: WST) detected a ransomware intrusion that encrypted systems globally and exfiltrated data. The company filed a material cybersecurity SEC 8-K on May 7, engaged Palo Alto Networks’ Unit 42 for incident response, and as of May 15 had restored core enterprise systems at some sites, with full restoration still in progress and financial impact not yet quantified.

What happened:

• West Pharmaceutical manufactures syringe components, vial closures, and drug delivery devices reaching approximately 70% of the world’s injectable drugs; no ransomware group had claimed credit as of May 15, which security researchers noted may indicate negotiated resolution
• This incident joins a three-event H1 2026 cluster: Stryker (March 11) and Foxconn (Nitrogen group, 8TB exfiltrated) all disclosed material cyber events in H1 2026, targeting globally distributed OT environments using double-extortion

Why it matters to you:

• West’s proactive global system shutdown is the correct containment playbook but delivers zero warning to customers; pharma manufacturers and CDMOs sourcing from West or comparable sole-source suppliers faced disruption with no prior notice
• The three-incident cluster confirms that third-party cyber risk programs limited to technology vendors are incomplete — tier-1 manufacturing and packaging suppliers must be assessed, and BCP scenarios must address 2–4 week supplier system outages

FDA’s 2026 Medical Device Cybersecurity Guidance Reaches Implementation Phase — Security Is Now Patient Safety Evidence

Practitioner communities actively published implementation analyses this week of FDA’s February 2026 cybersecurity guidance for software-driven medical devices — described as “a dramatic shift” in how FDA evaluates device safety and viability. The core change: cybersecurity is now a patient safety function, not an IT compliance requirement.

What happened:

• Cybersecurity architecture must now appear as an integral component of risk management documentation at pre-market submission — not a supplementary annex; post-market surveillance must include cybersecurity monitoring with triggers for reporting vulnerabilities as potential safety events
• The West Pharmaceutical and Stryker incidents illustrate the consequence: when enterprise IT and manufacturing OT are interdependent, encrypting one layer disables the other — the exact convergence risk FDA’s guidance requires medical device manufacturers to address structurally

Why it matters to you:

• PCCPs (Predetermined Change Control Plans), already required for AI-adaptive devices, must now incorporate cybersecurity change scenarios; IoT-connected devices and networked lab equipment face new firmware update, network segmentation, and vulnerability disclosure requirements
• CDMOs and pharma manufacturers operating MES and PAT systems interfacing with networked control devices must treat OT/IT governance as a unified risk domain under both FDA guidance and CISA’s OT guidance

🏢 Leadership & Operating Model

Three frameworks this week define what AI-era operating model leadership requires — from infrastructure through C-suite organizational architecture.

WEF/BCG: Life Sciences Has Shifted from Best Molecule to Best System — CIO Infrastructure Is the Foundation

A World Economic Forum report published May 14, developed with BCG and titled Strategic Choices in the Age of AI: Shaping the Future of Life Sciences, argues that competition has shifted from “who has the best molecule” to “who can operate the most effective system” — integrating data, experimentation, validation, and governance into a unified stack.

What happened:

• Data quality now matters more than volume; the ability to produce auditable evidence — tracing model evolution, data provenance, and analytical decisions — is as important as the result; regulation is moving toward governance embedded from the start
• The average upfront value of licensing deals between Western biopharma and Chinese organizations has risen more than 230% in recent years — with direct implications for how data sharing agreements and cloud infrastructure decisions are structured when Chinese-origin technology or collaboration is involved

Why it matters to you:

• Organizations that have not embedded traceability, version control, and auditability at the data infrastructure level will face increasing challenges as both regulators and partners raise expectations around continuous evidence generation
• The infrastructure investment question has shifted from “can we build this AI capability?” to “is our data architecture and governance infrastructure capable of supporting a continuously learning R&D system at regulatory-grade quality?”

2026 Is the Year Pharma Pilots End and Governed Agentic Workflows Must Begin

The defining challenge for life sciences IT leaders in 2026 is managing the transition from AI copilots to fully governed, production-grade agentic workflows that execute bounded units of regulated scientific work with defined outputs, provenance, and human oversight checkpoints.

What happened:

• Causaly’s framework defines three prerequisites for agentic deployment in pharma R&D: a specification upfront (intent, constraints, acceptance criteria), high-fidelity context (prior decisions, internal data, semantic memory), and a harness enforcing predictable output structure; without all three, systems cannot be validated in regulated environments
• ZS’s 2026 pharma survey found 41% planning to automate entire R&D discovery workflows with agents; KPMG’s 2026 Generative AI in Life Science survey found 63% mandate human validation of AI outputs — consistent with what FDA’s April warning letter now enforces

Why it matters to you:

• The shift from copilot to agent does not reduce the human accountability requirement — FDA’s Purolea warning letter demonstrates it intensifies it; deploying agents into GxP workflows without a formal governance framework creates the same regulatory exposure at larger scale
• Minimum viable governance before Q2 end: eligible workflow categories, specification requirements, CSA validation approach, human oversight checkpoints, and incident response for output failures

Lundbeck Names CAIO; BioMarin’s New CDIO Reports Directly to CEO — C-Suite AI Architecture Is Reconfiguring

Lundbeck appointed Markus Kede as SVP and Chief AI Officer — a dedicated C-level AI executive for strategy and deployment across R&D, manufacturing, and commercial. BioMarin Pharmaceutical established a new EVP, Chief Digital and Information Officer role and appointed Arpit Davé effective January 12, reporting directly to CEO Alexander Hardy.

What happened:

• Two distinct organizational models are emerging: BioMarin’s combined CDIO integrates data strategy, digital transformation, and IT infrastructure under a single executive with CEO access; Lundbeck’s dedicated CAIO manages AI strategy and governance separately from IT
• BioMarin created a new EVP-level position rather than elevating an existing CIO — the required capability profile is being redefined around executives who bridge scientific, regulatory, and commercial functions with direct board-level AI accountability

Why it matters to you:

• Both models reflect a common recognition: AI in life sciences is too consequential, too regulated, and too strategically central to be managed as a traditional IT function
• The dual-track CDIO + CAIO model is emerging at larger organizations with complex AI portfolios — a design decision with compliance implications given that AI governance in regulated workflows now requires executive-level accountability

💡 Editor’s Perspective

• The West Pharmaceutical attack and the FDA cybersecurity guidance are the same story from opposite directions. FDA’s February 2026 guidance requires medical device manufacturers to treat cybersecurity as patient safety evidence — West Pharma demonstrated exactly what that means when encrypting enterprise IT immediately disabled manufacturing for injectable packaging reaching 70% of the world’s injectable drugs.
• The Roche/PathAI acquisition and the AstraZeneca/Immunai expansion point in opposite vendor strategy directions. PathAI’s absorption into Roche removes an independent platform and creates conflict-of-interest risk for every non-Roche trial it supports; Immunai’s third renewal shows independent AI platforms with proprietary data assets remain viable — until an acquirer moves. Evaluate acquisition risk for every AI vendor in your clinical programs.
• The EU dual-compliance outcome and the FDA GMP AI warning letter define the same convergence point from different jurisdictions. Medical device manufacturers now face parallel AI Act and MDR/IVDR obligations requiring different documentation for the same AI system. FDA’s Purolea letter establishes that human accountability is non-delegable regardless of AI competence. Both push toward the same requirement: embedded governance from the start, not retrospective documentation.
• The agentic AI transition and the C-suite appointments are the same organizational design problem. Causaly’s framework requires specification, high-fidelity context, and an output harness before any agentic workflow can be validated in a regulated environment. BioMarin’s new CDIO and Lundbeck’s CAIO are organizational answers to who owns that governance architecture — organizations without both are creating the Purolea risk at larger scale.

🔗 Top 5 Must-Read Links

1. West Pharmaceutical Services: Company Impact Updates (SEC 8-K and operational status) — Primary source for the incident disclosure and restoration progress; essential reading for any CIO with supply chain cyber risk exposure to injectable packaging suppliers.
2. White & Case: EU Digital Omnibus AI Act Simplification Analysis, May 14, 2026 — The most comprehensive legal analysis of Digital Omnibus provisions and implications for medical device and pharma AI compliance programs; essential before updating your roadmap.
3. FDA Warning Letter WL 320-26-58 (Purolea Cosmetics Lab), April 2, 2026 — The actual warning letter establishing FDA’s enforcement position on AI in GMP document workflows; read the exact language before deploying generative AI in any regulated documentation context.
4. WEF/BCG: Strategic Choices in the Age of AI — Shaping the Future of Life Sciences, May 14, 2026 — The clearest articulation of the “product to system” shift in life sciences R&D; use it as the strategic lens for 2026–2027 infrastructure investment decisions.
5. Roche Press Release: Roche to Acquire PathAI, May 7, 2026 — Primary source for the acquisition announcement; biopharma organizations using PathAI for clinical trials should review contract terms and data governance provisions before close.

Every development this week reveals the same condition: the infrastructure choices you made in the last 18 months — vendor dependencies, data architecture, documentation quality, cyber segmentation — are being stress-tested simultaneously. If any of these threads connects to a live decision in your portfolio, hit reply — that’s what the community is for.

Ready to move beyond the digest? The LS CIO Community is where these conversations continue.

Join the LS CIO Community →

This digest is an interpretive summary of publicly available information and does not constitute legal, regulatory, cybersecurity, or investment advice.

Until next week,

Joe Miller

Founder, Leadership Inklings