New AI Agent Flags Cancer Immunotherapy Adverse Events
The research-stage tool identifies clinical safety signals, aiming to improve consistency in oncology review processes.
Updated on Oct. 29, 2026 in Artificial Intelligence
Researchers at the University of Lausanne and Heidelberg University Hospital have developed an agentic AI system designed to identify immune-related adverse events in cancer immunotherapy clinical notes. The research-stage system uses a large language model to scan prospective clinical documentation for conditions that are often difficult to detect during routine practice.
Why it matters
Identifying immune-related adverse events is a significant challenge in oncology care, as these reactions are frequently missed in routine clinical practice. Automating this detection process could improve patient safety by ensuring these events are flagged more consistently and rapidly than manual review allows.
The system employs an agentic AI framework—software designed to perform autonomous tasks—powered by a large language model to process unstructured clinical text. It is specifically tuned to identify complex immune-related adverse events, which are inflammatory responses triggered by cancer immunotherapies.
The players
University of Lausanne
A research institution in Switzerland focusing on biomedical innovation and clinical outcomes.
Heidelberg University Hospital
A major German medical center that conducts clinical oncology research and develops digital health tools.
The details
The tool functions as an autonomous agent that navigates clinical documentation to isolate potential safety signals. By leveraging large language models—algorithms trained on vast text corpora to interpret and generate human-like language—the agent extracts patterns associated with adverse drug reactions. This approach addresses the limitation where subtle clinical indicators are overlooked during standard, manual record review.
Timeline
October 29, 2026: The research describing the adverse event detection tool was published.
The Tech Race
This development aligns with the broader push to implement agentic AI in clinical oncology, moving beyond simple diagnostic image analysis. It competes with other automated medical record review systems currently being tested to reduce the burden of manual safety reporting in high-volume hospitals.
This AI tool is currently in the research stage and is not yet available for general clinical use in hospitals. Its eventual deployment depends on validation studies to determine how it integrates with existing electronic health record systems and oncology clinical workflows.
The takeaway
This research highlights the shift toward using autonomous AI agents to solve specific, high-stakes clinical diagnostic challenges. Watch for future peer-reviewed reports on the system's error rates compared to human oncologists in large-scale clinical trials.
Further reading
For more on the development of AI systems in medicine, see our coverage in Artificial Intelligence.
Source note: This article includes information reported by Nature.
