AI for Oncology
Reply and IEO Collaborate to Develop Specialized AI Models for Oncology
Reply and the European Institute of Oncology (IEO) have launched a partnership to build cancer-specific AI models — a move that could directly affect the 18,000+ patients IEO treats each year, according to Business Wire. The collaboration will train Large Language Models, or LLMs, on real clinical data from one of Europe's top cancer hospitals.
Unlike general AI tools like ChatGPT, these models will be built on IEO's own patient data and medical knowledge. The first targets are breast cancer, urologic cancer, and prevention, Business Wire reported.
Reply Model Factory Powers the IEO Partnership
The project runs on Reply's newly launched platform, the Reply Model Factory, which went live on May 14, 2026. Think of it as an assembly line for building AI — not one-off experiments, but production-grade systems built around each organization's own data. Carlo Malgieri, a Partner at Laife Reply, put it plainly: "Real value will come from models built on each organization's own knowledge, data, and expertise," according to Business Wire.
Reply recorded €2.45 billion in revenue for 2025, a 6.7% increase driven largely by its AI push. A successful rollout at IEO would give the company a blueprint to sell similar AI systems to hospitals across Europe.
Why Generic AI Falls Short in Oncology
General-purpose AI models are trained on public web data. That creates a problem in medicine. Experts writing in the JCO Clinical Cancer Informatics series found that these models often produce errors — called "hallucinations" — or miss critical gaps in oncology knowledge. Domain-specific models, trained on real clinical records, show up to 94.7% concordance with expert medical boards in similar research settings.
Annarosa Farina, IEO's Director of Information Systems, said: "AI is not simply a technology; it is a valuable ally to medicine — helping us to interpret the complexity of cancer through the analysis of large volumes of clinical and scientific data." IEO currently runs 698 active clinical trials and holds 307 hospital beds in Milan, giving the project a rich data foundation.
This Partnership Builds on Years of IEO-Reply Work
This is not the first time Reply and IEO have worked together. In July 2025, the two launched X-RAIS — an AI agent that helps radiologists classify breast cancer lesions in real time during mammography screening, according to Business Wire. Before that, in late 2024, IEO kicked off the BIANCA Project, a 36-month effort to build Italy's first AI-powered digital biobank, funded under Italy's Ministry of Enterprises with €731 million in national innovation grants.
Strict Rules Govern Healthcare AI in Italy and the EU
The project must operate inside a tight legal box. The EU AI Act, passed in 2024, classifies healthcare AI as "high-risk." That means strict data rules, human oversight at every step, and full transparency. Italy added its own layer in September 2025 with Law No. 132/2025 — the country's first national AI law. Italy's data regulator, the Garante, has already fined OpenAI €15 million for data violations, signaling zero tolerance for missteps.
Italy's national AI strategy also pushes for "strategic autonomy" — keeping Italian health data under Italian control, not processed by closed foreign systems. That political pressure is one reason Reply is building its own Model Factory rather than using off-the-shelf US tools. The next phase of the IEO project will train LLMs on specific clinical use cases to support prevention, diagnosis, and treatment, according to Business Wire.
Sentiment
NeutralPublishers
4
Articles
3
Reach
4
