How the map
decides what
to show.

We grade every incident on the scale used by the World Health Organization's Surveillance System for Attacks on Health Care:

The grade says how sure we are that the event happened. It does not say that every detail is right. Early reports of disasters and attacks are often wrong about numbers and places, and figures change as more is known. When a later report gives a higher figure for an incident we have already shown, the incident is updated and marked as such.

Some further rules: only an official source or an independent outlet can move an incident to Confirmed or Probable; the AI model can never do that on its own. Incidents drop off the map after seven days unless new reports keep arriving, and nothing is kept longer than 60 days. Each month we count how many WHO Disease Outbreak News items we had a report on first, and how many single-outlet reports were later backed up, and we publish that scorecard here when there is a full month of data.

Included: current events that affect people's health or safety and are reported by a named source: disease outbreaks and unexplained illness; earthquakes, tsunamis, storms, floods, landslides, volcanoes, wildfires and drought; riots, unrest and armed conflict; famine and hunger; and attacks on health care, meaning violence against, or obstruction of, health workers, patients, facilities, ambulances and supplies. For attacks on health care we use the definitions in the WHO Surveillance System and the Safeguarding Health in Conflict Coalition methodology as our reference, and we point to their datasets rather than compete with them.

Not included: films, series, books, games, exercises and simulations; sport and entertainment; opinion, analysis and explainer pieces; anniversaries and historical pieces; figures of speech ("a tsunami of complaints"); reports that give no place; reports from state-run media about countries other than their own; anything from social media, messaging channels or anonymous accounts; and any report we cannot link to.

Coverage is uneven. Our feeds are densest in the countries where we work or plan to work (DR Congo, Uganda, Nigeria, Sudan, Chad, Ukraine and their neighbours) and thinnest where little is published online. A gap on the map means we found nothing reportable, not that nothing happened.

Bubbles are placed at the centre of the country, state or province, not at the event. The number is how many reports the incident has. Red, orange and green rings follow the official agency's own severity rating where there is one.

Information only. This service is automated and partly machine-translated. It is not a substitute for advice from the Foreign, Commonwealth & Development Office, the World Health Organization or local authorities, and it must not be used for safety-of-life decisions. We give no guarantee that it is complete, accurate or up to date. It is not medical, security or travel advice, and it is not a duty-of-care or emergency service.

The monitor will sometimes misread a story, put it in the wrong place or miss it altogether. In our own tests the translation model made a meaning-changing error in roughly one headline in fifty to a hundred, which is why every translated line is labelled and linked to its original. No single system catches every event: in one published evaluation of seven epidemic-intelligence systems, each caught between 38% and 72% of the influenza events studied.

Always check the original source and official advice before acting on anything you see here.

We follow the OCHA Data Responsibility Guidelines and the Signal Code as far as they apply to a small automated service:

• No personal data on the map. We publish no names or details of victims, patients, health workers or witnesses, and our summaries are written to leave them out.
• No exact locations of people or facilities. Incidents are shown at country or province level. In active conflict we will not show a finer location even when a source gives one.
• Group risk counts too. Where a report identifies a clinic, camp or community by the group it serves, we keep the summary general.
• A harm check before each new data layer. Before we add a new source or type of incident we ask who could be put at risk by its publication and whether to delay, blur or withhold it.
• Nothing about you. We don't collect information about the people who use the map. See our privacy notice.

If a report on the map is wrong, misleading, in the wrong place, or puts someone at risk, email contact@aeiaxioma.uk with the link or a description. We aim to look at every request within two working days, and to remove anything that could put people in danger straight away. Corrected incidents are marked as corrected rather than silently changed. Publishers who would rather we didn't link to their reports, and anyone who believes a summary misrepresents them, can ask the same way.

Our summaries are licensed under CC BY 4.0: credit "Axiom Global Medicine incident monitor" with a link to axiomglobalmedicine.uk. Each report links to its original publisher, whose content stays theirs. Developers can use the same data through our incident data API.

Official sources can confirm an incident on their own. News outlets can make it probable when two or more agree. Rumour sources can never make it public. The terms below are our reading of each publisher's published licence or terms of use as at 31 October 2026; tell us if we have one wrong.