M.A.I S.I.G.U.RPublic Safety Map
Romania's crime map · Bucharest pilot

M.A.I S.I.G.U.R
See · Understand · Prevent

Sistem Inteligent de Geoanaliză Urbană a Riscului (Intelligent Urban Risk Geoanalysis System) - Romania's first public safety map. We gather official data, anonymise it and analyse it with artificial intelligence, to bring to light what is hard to see and to support safer communities: residents, city halls and the Poliția Locală (local police).

Open the map → Why we are doing this
Official DGPMB data by sector
Anonymised by design
GDPR & AI Act compliant
Crime coefficient · sector 6 sectors, 6 years
Explore
Built on official sources
Why we are doing this

Helping you build a safer community

In Romania, crime data exists - but it is fragmented, aggregated at county level and published as static reports, with no map and no context. MAI SIGUR ("safer" in Romanian) brings it together, puts it on a map and analyses it with artificial intelligence that learns from patterns and estimates where risk may concentrate - turning it into public infrastructure for transparency and prevention.

Inform residents

An informed person makes better decisions: which route to take, where to live, how to protect their family. Awareness is the first step of prevention.

Rebuild trust

Full transparency is compatible with data protection. A public map, aggregated and anonymised, narrows the distance between the community and the authorities.

Guide local decisions

City halls and the local police see where to direct resources: lighting, patrols, surveillance. Data for better public decisions.

The context

A functional gap and an officially recognised need

Romania's National Public Order and Safety Strategy 2023-2027 (SNOSP) explicitly calls for regular measurement of how safe communities feel. The figures below come from official sources.

Figures from SNOSP 2023-2027 (Government of Romania) and ARSC analysis of data.gov.ro. Full details and sources on the Methodology page.

The model

We anticipate risk by area, we do not profile people

We analyse places and patterns, never people. We estimate where risk may concentrate in order to support prevention, but we do not track or profile individuals - that would be "predictive policing", which the AI Act prohibits. Three complementary layers make up a risk map that city halls can use for concrete decisions.

Read the full methodology →
01

Aggregated official data

The official crime coefficient by sector (DGPMB) and public statistics, geolocated and anonymised by anchoring to an area. In later stages, integration with the local police and IGPR/DGPMB.

02

Civic reports

Residents report how safe places feel: broken lighting, abandoned areas, dangerous junctions, antisocial behaviour. This answers the SNOSP requirement directly and builds civic engagement.

03

Urban vulnerability

An overlay of infrastructure: CCTV coverage, road accident black spots, street lighting, proximity to schools. Context that turns data into prevention.

Artificial intelligence & compliance

Data aggregated with AI, behind the scenes, for responsible predictions

A large language model (LLM) cleans, standardises and classifies the data, identifies patterns and estimates concentrations of risk by area - not by person. Processing runs on European infrastructure (Mistral AI), to comply with the GDPR and keep the data within the EU.

// 01

Anonymised by design

No identifiable personal data. Incidents are anchored to the nearest area, not the exact address - following the police.uk model validated by the data protection authority.

// 02

Aggregation with a minimum threshold

Data is shown only above a minimum number of events per area. Below that threshold, nothing is shown - so that re-identification is impossible.

// 03

Sovereign European AI

Processing through Mistral AI (EU), not through services outside Europe. The data stays in the EU, in line with GDPR requirements on data protection and transfers.

// 04

Place-based, not person-based

The analysis looks at places and patterns, not individuals. No arrest data, no profiling - we avoid discriminatory "feedback loops" and stay within the letter of the AI Act.

Legal framework

Compliant with the GDPR (Reg. 2016/679) and Legea 190/2018 (Law 190/2018), Legea 363/2018 (Law 363/2018) (processing of data by law enforcement authorities), the EU AI Act (aggregated, place-based analysis - permitted) and Legea 179/2022 (Law 179/2022) on open data.

Reference models

What Western police forces have built - and what we are adapting for Romania

police.uk

The gold standard of transparency. Monthly open data, at street level, 14 categories, anonymisation designed together with the data protection authority.

UK AI Crime Map

The 2025-2030 generation: an AI-powered predictive map that anticipates concentrations of risk. The direction of travel - from reaction to prevention.

CrimeMapping / LexisNexis

The American model: interactive maps plus alerts for residents on "areas of interest". The strongest driver of civic engagement.

FixMyStreet

The citizen participation layer: simple civic reports, on very safe legal ground. Our model for Layer 2.

The prediction feature

See where risk may concentrate - before it happens

On the map, switch on the "AI prediction · 30 days" layer. Pulsing hotspots appear, each with an estimated risk score (0-100) for that area. The model does not predict individuals or specific offences - it estimates the concentration of risk by place, so that city halls and the local police can direct resources preventively.

A

What goes into the model

The recent density of events per area (90 days), the sector's official DGPMB coefficient and its trend compared with the previous year.

B

How it is processed

An LLM (Mistral AI, European infrastructure) cleans, classifies and weights the data. The resulting score is explainable - each hotspot's card shows what it is made of.

C

Where it is used

In planning patrols, lighting and surveillance. In line with the AI Act: place-based, aggregated, never individual profiling.

See the prediction on the map →
Risk prediction · 30 daysEstimated hotspots
Bucharest pilot

See how safe your city is, on a map

Explore the official crime coefficient for Bucharest's 6 sectors and the demonstration layer of incidents. Switch on alerts for your area of interest.

Open the map → Set up alerts

The Smart City Romania network, run by ARSC

Everything you need to know about Romania's fastest-growing industry

Explore

Take part

Read

arsc.ro · [email protected]