Responsible AI

How AidGrid uses AI — and how it doesn't.

AidGrid AI triages and suggests. Human coordinators approve every routing decision. No automatic dispatch.

What AI does in AidGrid

  • Reads every request
  • Scores urgency (1–5) based on keywords, context, and patterns
  • Catches duplicates (same address, similar description, same source)
  • Matches resources (cross-references requests against the resource directory)
  • Translates (requests in other languages → English)
  • Organizes (sorts and filters the coordination queue)
  • Flags concerns (incomplete submissions, unusual patterns, suspicious information)

What AI does NOT do

  • AI does not confirm that a person is actually in danger
  • AI does not decide who gets help — humans decide
  • AI does not replace emergency services (911 or local equivalent)
  • AI does not automatically dispatch anyone anywhere
  • AI does not diagnose medical conditions
  • AI does not verify resource availability — it suggests, coordinators confirm
  • AI does not replace field coordinators, shelter managers, or response teams
  • AI does not share your personal information with other organizations

For life-threatening emergencies, contact local emergency services (911 in the US, 112 in the EU). AidGrid is a coordination layer for response teams — not a dispatch service.

Scope & non-claims

The AidGrid scope, in plain language

AI suggests. Human coordinators approve.

AidGrid can

  • Receive and organize incoming help requests
  • Score urgency as a suggestion for coordinators
  • Flag potential duplicate reports
  • Suggest nearby resources from the directory
  • Show capacity signals from shelters and clinics
  • Alert responders when a case needs review
  • Export case records for reporting and handoff

AidGrid cannot

  • Guarantee a response
  • Verify every report
  • Dispatch responders automatically
  • Decide who gets help
  • Replace emergency services
  • Diagnose medical needs
  • Provide legal guidance

Not intended for

  • Emergency dispatch or a 911 replacement
  • Medical diagnosis or treatment
  • Legal advice
  • Direct public fundraising

For life-threatening emergencies, call 911 or your local emergency number first.

Human review — always

"Every routing decision — assigning a team, referring to a partner, marking resolved — requires a human coordinator to approve. The AI prepares the information. A human makes the call."

How human review works

  • Coordinators review every AI triage output before routing
  • Volunteers verify resource capacity and translation quality
  • Partners review updates before wide distribution
  • Community members can report corrections at any time
  • All routing decisions are logged and attributed

AI boundary by scenario

Different responses, same boundary.

Flood response

AI can suggest urgency and catch duplicates. A coordinator confirms routing to rescue teams.

Shelter capacity

AI matches requests to available resources. Shelter managers confirm availability before people are routed there.

Medical requests

AI scores urgency. Qualified responders and emergency services handle medical decisions.

Training and improvement

  • AI improves based on coordinator feedback
  • Volunteers review translation quality and flag errors
  • Field feedback shapes triage accuracy over time
  • No AI decision is final without human review
  • Regular human review of AI outputs for bias and accuracy

Reporting an AI error

Think the AI got something wrong? Report it.

Every report is reviewed by a human. Field feedback is how the system improves.

Report an AI errorsafety@aidgrid.org