Skip to content
This is NOT an official site of the City of New York. Click here for official AI reports.

NYC Office of Technology and Innovation

NYC - DTPR AI Register

Back to your search

AI Risk Scoring for Fire Inspection Prioritization

What it is for

Risk Assessment & Triage
  • ALARM scores each of approximately 330,000 buildings on a one-to-nine risk scale (one = highest risk) combining a machine-learning probability of structural fire ignition with a civilian injury/death risk ratio, used to prioritize Fire Department operational inspections.
Safety & Security
  • The system supports preventive fire safety by ensuring that the highest-risk buildings in New York City are inspected first, reducing the probability of structural fires and civilian casualties.
Run by
Fire Department (FDNY)
Where
No fixed location
Kept
Retained 5 years (training dataset window); scores updated daily
Shared with
Accountable organization
Your copy
You cannot see the data it holds about you. What you can do

What it is for

Risk Assessment & TriageSafety & Security

This system scores every inspectable building in New York City (approximately 330,000 properties) on a scale of one to nine, where one indicates the highest fire risk. The score combines a machine-learning estimate of structural fire probability with a risk ratio for civilian injury or death, and is used by the Fire Department to schedule which buildings receive operational inspections. No identifying information about individual occupants is collected or disclosed; the system affects properties and, by extension, the people who live or work in them.

What it collects and what happens to it

Data taken in

Operational data
Anonymized data
  • Fire and Emergency Medical Services dispatch records, building characteristic data, 311 call data, felony crime data, census data, and civilian injury data — all used as aggregate operational inputs to compute building-level risk scores. No identifying information about individual people is collected or used.

Processing

Classification & Prediction
  • A machine learning model trained monthly on a five-year incident dataset (99% train split for ignition model, 80% for impact model) predicts structural fire ignition probability. This is combined with a risk ratio for civilian injury/death to produce a composite risk score.

What it does

Deciding (Analytical AI)
Human decides
  • The system produces a risk score (one to nine) for each building, which is used to schedule inspections. Fire operations staff use the score to prioritize inspection schedules; the score informs but does not automatically execute any inspection or enforcement action.

Outputs

A recommendation or prediction
Anonymized data
  • A risk score from one (highest risk) to nine (lowest risk) is output for each building. Buildings within the same score tier are further ranked by residential population. This score is an advisory prioritization for inspection scheduling, not a binding enforcement determination.

Run by

Fire Department (FDNY)
  • The NYC Fire Department's Management, Analysis, and Planning unit deploys and operates ALARM/RBIS to schedule fire operations building inspections across approximately 330,000 properties citywide.

NYC Algorithmic Tools Compliance Report

Built by

Not stated by the NYC Office of Technology and Innovation.

Kept for

Retained 5 years (training dataset window); scores updated daily
  • The model is retrained monthly using a rolling five-year incident dataset. Risk scores are recalculated daily. The register does not specify an explicit retention period for output scores.
  • Duration: 5 years (training dataset window); scores updated daily

Shared with

Not available to me
  • Individual building risk scores are not disclosed to members of the public or to building occupants. The system is used internally by Fire Department operational staff for inspection scheduling only.
Available to the accountable organization
  • Risk scores and inspection scheduling outputs are available to NYC Fire Department operational and planning staff for the purpose of scheduling fire inspections.

Stored

Stored locally
  • The system is operated by the NYC Fire Department, a city agency. Data storage location is not explicitly stated in the register, but as an internally developed city government tool with no external vendor, local storage is the most reasonable inference.
  • Duration: Not specified; 5-year rolling training window confirmed
How to read the colours

Can it identify you?

Anonymized data
Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
Pseudonymous data
Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
Identifiable data
The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.

Who completes the loop?

Human decides
This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
Human executes
This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
Autonomous
This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.

Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.

What you can do

Ask about this system

Questions go to the NYC Office of Technology and Innovation, not the vendor.

Your rights

  • Right to Be Informed of AI UseBuilding owners and occupants are not individually notified that their property has been scored or selected for inspection by this AI system. The existence of the ALARM/RBIS tool is publicly disclosed through the NYC Algorithmic Tools Compliance Report, available at the NYC Open Data portal.
  • Right to Algorithmic TransparencyThe general methodology of ALARM — combining a machine learning ignition probability with a civilian risk ratio to produce a one-to-nine score — is disclosed in the NYC Algorithmic Tools Compliance Report. The specific model architecture, feature weights, and full training data schema are not publicly available.

Risks and safeguards

  • Civil liberties harmThe inclusion of nearby felony crime data in the civilian impact score could embed geographic or demographic bias, potentially over-inspecting buildings in communities that already face disproportionate policing.Safeguard: The register states no identifying information is collected or disclosed; scores apply at the building level, not to individuals. The model is retrained monthly on updated data, which allows bias patterns to be monitored and corrected over time.