Last month, Doctors Without Borders/Médecins Sans Frontières (MSF) announced an award of US$1.3m from the Google Artificial Intelligence Impact Challenge to its innovation entity, the MSF Foundation, in order to develop a smartphone app to help AMR diagnosis in low-resource settings. The app – titled ASTapp – assists in interpreting antibiograms (tabular or pictorial representations of antibiotic susceptibility tests) in settings with limited access to expert staff and microbiological tools.

This project enters an increasingly diverse landscape of projects aimed at applying machine learning, automated pattern recognition, and computer-mediated decision support in the field of health care, such as the Google Ventures-backed Okwin, which is developing a tool to enable pharmaceutical companies to trawl one another’s data to accelerate drug development, using blockchain technology to ensure traceability and protect commercial secrets, and the Australian OUTBREAK ‘knowledge engine’, mapping AMR fingerprints across people, animals, and the environment.

To the ASTapp project, Google brings the grant funding, plus “credit and consulting from Google Cloud, mentoring from Google AI experts and the opportunity to join a customized accelerator program from Google Developers Launchpad.” The MSF Foundation is partnering with France-based academics at Evry and Créteil, and brings its own “foresight and audacity” to the task of improving the effectiveness of humanitarian field medicine.

In this crowded and well-funded research space, it becomes increasingly important to develop means of ensuring that new technologies don’t merely reproduce resource inequalities. The technical outcomes will surely be subject to stringent audit and review, as will any academic outputs. Prominent attention should also be paid to the widest possible diffusion of technical capacity (to originate and scrutinise projects), and intellectual property rights, to ensure that the stakes in managing the risks and developing, using, and sharing the benefits of data analysis accrue in traditionally low-resource settings, which are used to bearing the costs of new and growing risks, with minimal systemic gains from benefits.