SafeZone Map colors neighborhoods across 15 countries by how much crime is publicly reported there, normalized by population. When there isn't enough data, the map says so — gray means unmeasured, never "safe."
Zones are ranked against their peers by reported-incident levels per resident over a rolling window. Red is "highest relative to comparable areas" — not a verdict, and not an official crime rate.
Lower reported-incident levels than most comparable areas.
Middle of the range for its cohort.
Highest reported-incident levels relative to peers.
Insufficient data. Unmeasured is not the same as safe — we never conflate the two, and we think this honesty is the whole point.
News reports and official open data become anonymous incident records: each report is checked for relevance, reduced to facts (type, place, date), located to a neighborhood, and de-duplicated so three articles about one incident count once. Severity comes from a fixed lookup table, never a model's opinion. Then, for every zone:
Old incidents fade with a six-month half-life; official sources count more than tabloids; and the result is bucketed by percentile against comparable areas — so red always means "worst relative to peers," never an absolute claim.
Severity comes from a fixed lookup table, never the model's opinion. Article text is discarded at extraction.
Recent reported incidents appear on the map as they are processed, hour by hour.
Compare two neighborhoods side by side, save up to 20 favourites on your device, and see what changed since you last looked.
A built-in assistant answers questions about any covered area — grounded only in SafeZone Map's own data, with the same honesty rules.
Each incident links back to the original publisher's article or official dataset.
No accounts, no profiles, no people's information at rest. Article text is processed and immediately discarded in the same operation that stores the anonymous incident — what remains is a type, a severity, a date, a place, and a link to the publisher's own copy. Your saved places live only in your browser. Analytics runs only if you opt in.
Neighborhood-level percentile scores, score history, and cited incident facts across all 15 countries — structured, PII-free, and key-authenticated. Built for proptech, insurance, relocation, and newsroom use cases.
No. SafeZone Map measures relative levels of publicly reported incidents, normalized by population. Official statistics seed baselines where they exist at neighborhood granularity, but the map's claim is always "reported incidents relative to peers."
Gray means there isn't enough data to score the area honestly — too few reported incidents in the window, no population baseline, or no neighborhood-level source yet. We show that plainly instead of guessing. Gray is not good news or bad news.
Public news reporting (processed hourly, every incident cited and linked to the original publisher) and official open datasets such as national statistics offices and police open data. Sources are weighted: official data counts fully, tabloids count least.
The design guards against it: scores are relative percentiles with citations, no area is ever labelled "safe" or "dangerous," the disclaimer travels with every view and share card, and unmeasured areas are gray rather than assumed anything.
Open the free map, or tell us what you're building — we'll point you at the right integration.