News
Sanofi and Owkin expand multi-year collaboration to build agentic AI for pharmaceutical R&D
Owkin and Sanofi announced a multi-year collaboration backed by a five-year license for K Pro, Owkin's AI Scientist platform. The partnership extends a €90M collaboration that began in 2021 with oncology target identification. AI agents built under the new program will autonomously perform complex drug research and development tasks — competitive intelligence, patient subgrouping, drug positioning — drawing on K Pro's multimodal patient data fusion capability. Initial focus is oncology and immunology pipelines.
- pharma
- drug-discovery
- agents
- agentic-AI
- oncology
The Sanofi-Owkin collaboration is a useful case study in how pharma companies are structuring their agentic AI bets differently from their earlier LLM pilots. The 2021 original partnership was a data-science-as-a-service arrangement: Owkin’s federated learning platform ingested Sanofi’s multimodal patient data and generated statistical models for target identification. The 2026 expansion is architecturally different — it is asking Owkin’s AI agents to autonomously execute research workflows, not just generate outputs that scientists review.
K Pro, Owkin’s AI Scientist platform, is designed to fuse multimodal patient datasets — genomics, histopathology, clinical records — with biological AI models specialized for pharmaceutical workflows. The “agentic” framing means the system can chain together subtasks: pull competitive intelligence from literature, cross-reference it with internal target data, identify patient subgroups in trial datasets, and recommend drug positioning without a scientist manually orchestrating each step. Whether it delivers on that description in practice is the implementation question, but the contract structure (five-year K Pro license) suggests Sanofi is betting on K Pro as infrastructure, not a one-off project.
This deal fits inside a broader pharma-AI agentic pivot in 2026. Pharma companies that signed LLM partnerships in 2023-2024 for literature summarization and meeting preparation are now asking what comes next — and the answer is consistently “more autonomous task execution, less human routing.” The risk is that autonomous research workflows introduce failure modes that are harder to catch than in supervised settings: an agent that makes a plausible but wrong inferences about a drug-target relationship will propagate that error into downstream experiments before a scientist notices.
The oncology and immunology scope reflects where Owkin has the most depth in multimodal patient data. Owkin’s federated learning history gave it access to pathology slide repositories across European hospital networks that few other AI companies can match. For Sanofi’s pipeline, oncology and autoimmune indications are both heavily dependent on patient subgrouping and companion diagnostics — exactly the problem Owkin’s architecture was built to address.
Related coverage: health-system copilots topic.
Primary source: Read the full original on BusinessWire ↗