Antibiotics, AI and the Slow Emergency

Some emergencies announce themselves. Antimicrobial resistance often does not. It accumulates in hospitals, farms, clinics, wastewater and supply chains. A treatment works a little less often. A routine infection becomes complicated. Surgery, chemotherapy and intensive care become more dangerous because the protective background of effective antibiotics weakens.

A miracle taught us a warning

Penicillin transformed medicine, but Alexander Fleming warned almost immediately that misuse could train bacteria to resist it. That is the uncomfortable history of antibiotics: every triumph contains a timer. Resistance is not a failure of science; it is evolution responding to human behaviour.

WHO describes antimicrobial resistance as a condition in which infections become difficult or impossible to treat. CDC data tie the problem to at least 1.27 million deaths worldwide in 2019 and nearly 5 million associated deaths. These numbers are not a future scenario; they are a present burden.

Drug resistance can make infections impossible to treat.

AI can help discover compounds, predict resistance patterns and improve prescribing. But the slow emergency will not be solved by discovery alone. Stewardship, sanitation, vaccination, diagnostics, livestock policy and access to second-line medicines are as important as the next model.

AI may help in several ways: searching chemical space, identifying resistance patterns, supporting diagnostics and improving antibiotic stewardship. But AMR is not only a discovery problem. It is a behaviour problem, a market problem and a global-coordination problem.

New antibiotics are difficult to commercialise because the best public-health use is often restrained use. A drug that should be held in reserve is not a normal blockbuster. Diagnostics may exist but not be used. Stewardship may be praised but underfunded. Low- and middle-income health systems may face both overuse and lack of access, a paradox that cannot be solved by technology alone.

The slow emergency requires a different moral imagination. Society is better at mobilising for visible catastrophe than for the preservation of background conditions. Effective antibiotics are a public good: unnoticed when present, terrifying when absent.

AI can be part of the answer, but only if paired with incentives, surveillance, access and restraint. The goal is not simply to discover more weapons. It is to preserve the conditions under which ordinary medicine remains ordinary.