The European Commission’s Joint Research Centre (JRC) says generative AI could significantly improve disease surveillance by quickly extracting and organising information scattered across outbreak reports, news articles and other unstructured sources. Researchers tested AI prototypes using thousands of reports from the WHO’s Epidemic Intelligence from Open Sources system and used them to build an epidemiological knowledge graph linking information from different sources.
One of the biggest potential benefits is speed. The JRC found that AI could synthesise large volumes of epidemiological information much faster than manual review, potentially shortening the gap between detecting an outbreak and responding to it. Researchers also used retrieval-augmented generation over WHO Disease Outbreak News to create more detailed health-threat narratives, which could improve situational awareness and preparedness for public-health authorities.
However, the technology is still experimental rather than operationally proven. The JRC notes that the prototypes have not yet been validated in real-world public-health settings, while large language models can generate convincing but incorrect information. Bias, privacy and individual-rights concerns are also particularly serious in disease surveillance because an incorrect interpretation could influence decisions about an emerging health threat.
The broader takeaway is that AI should act as an early-warning and analysis layer, not as an autonomous public-health decision-maker. The JRC recommends phased national or regional pilots, strong validation and bias-mitigation mechanisms, interoperable data standards and continued human oversight before wider deployment. The approach reflects a broader principle emerging across high-stakes AI: use machines to process information faster, but keep responsibility for consequential decisions with qualified humans.
