UN agencies set guidelines for health AI innovation

WHO, ITU and WIPO launched joint strategic guidelines at the AI for Good Global Summit in Geneva, offering health innovators a roadmap for intellectual property, data governance and regulatory pathways. The non-binding framework recommends combining patents for technical methods with trade secrets for proprietary datasets, while maintaining quality assurance, safety monitoring and patient privacy.

The guidance emphasizes equity by design through field-of-use licensing, differential pricing and tiered cloud access. These models could allow developers to preserve commercial markets while enabling technology transfer and subsidized access in lower-income regions. Speakers warned that voluntary standards must be translated into national rules, especially as health systems turn to AI for efficiency under fiscal pressure.

Examples highlighted in Geneva included AI-assisted liquid biopsy tests that can detect multiple cancers at stage one, when survival rates reach up to 92%, and a pilot tool using cough sounds to identify tuberculosis. Global Fund officials noted that TB programs still miss something like 3.6 to 4 million people a year, underscoring the appeal of scalable diagnostics over basic mobile networks.

Health leaders also warned that imported models trained on unrepresentative data can deepen disparities. Approximately 90% of global genomic data currently belongs to people of European descent, raising risks for predictive tools deployed elsewhere. The guidelines call for local validation, shared performance data across populations and regulatory cooperation to build trust in medical AI.