evidoria

← Back to browse

Good practice Imported

Aduana Inteligente — Brazil's Receita Federal Screens Airport X-Ray Images With Google Gemini

Brazil · Guarulhos · See the Brazil profile

Evidence: Observational / pre–post Top 67% 53/100 · Ask Evidence Copilot about this practice

Since December 2024 Brazil's Receita Federal has fed Guarulhos Airport's X-ray feed into Google's Gemini model, flagging suspect baggage and parcels in real time and cutting international-purchase delivery times from up to two months to under a week.

800,000 shipments/day
Average daily international shipments received by Brazil (as of 2025)
under 1 week
Reported customs clearance time for AI-screened shipments (down from up to two months) (since Dec 2024 rollout)
Aduana Inteligente — Brazil's Receita Federal Screens Airport X-Ray Images With Google Gemini

Details

Maturity
Pilot
Promoter
Receita Federal do Brasil, with Google Cloud
Period
2024-2026
Keywords
customs, tax administration, border security, logistics

Context

Since December 2024, Receita Federal (Brazil's federal tax and customs authority) has integrated Google Cloud's Gemini generative-AI model directly into the X-ray scanning line at Guarulhos International Airport in São Paulo, Brazil's busiest air-cargo gateway.

Activities

As baggage and international parcels pass through the scanner, Gemini reads the image in real time and automatically alerts a customs officer when it detects an irregularity — undeclared goods, weapons, vehicle parts, medicines or drugs — rather than routing every item for manual inspection, letting officers concentrate manual checks on the small share of items the model flags as high-risk.

Results

Brazil receives an average of roughly 800,000 international shipments a day. Before the system, imported purchases could take up to two months to clear customs; Receita Federal and Google Cloud report that with AI-assisted screening, some shipments now arrive in under a week.

Conclusions

The reported delivery-time improvement comes from the implementing agency and vendor rather than an independent audit; no published false-positive or false-negative rate, appeals mechanism, or algorithmic impact assessment accompanies the rollout, and the system remains limited to a single airport rather than Brazil's full customs network.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Medium (1–3 years)
Staffing & skills
Receita Federal (Brazil's federal tax and customs authority) operates the screening line, with Google Cloud providing the Gemini model integration, Customs officers perform manual inspection only on items the model flags as high-risk

Conditions for success

  • Real-time AI reading of the existing X-ray scanning line at Guarulhos, rather than a separate inspection step, lets officers concentrate manual checks on flagged high-risk items

Common failure modes

  • Reported delivery-time improvement comes from the implementing agency and vendor rather than an independent audit
  • No published false-positive/false-negative rate, appeals mechanism, or algorithmic impact assessment
  • Deployment remains limited to a single airport rather than Brazil's full customs network

Where it fits

Governance type
national tax/customs authority with private technology vendor partnership
Scale
single site (Guarulhos Airport)
Income level
upper-middle-income

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful