CBSA Machine Learning System Scores All Border Entrants with No Independent Audit
CBSA's Traveller Compliance Indicator assigns compliance scores to all border entrants at land ports, expanding nationally by 2027, with no published Algorithmic Impact Assessment, no reported independent audit, and expert-identified bias concerns.
The Canada Border Services Agency has deployed a predictive analytics tool called the Traveller Compliance Indicator (TCI) that assigns a compliance score to travellers entering Canada at land border ports of entry. The system is built on five years of traveller compliance data and is intended to direct officer attention to higher-risk entrants. The actual decision on whether to refer a traveller for secondary examination rests with the border services officer.
The TCI was piloted at six land ports of entry in 2023. In September 2025, reporting revealed that CBSA plans to expand the system to all land ports of entry by end of 2027, with air and marine ports to follow. CBSA confirmed the expansion timeline.
University of Toronto professor Ebrahim Bagheri, a responsible AI researcher, identified a "major risk" of bias "against certain subpopulations" — citing patterns consistent with bias documented in other AI risk-assessment systems, such as the COMPAS recidivism tool. Bagheri stated: "The only way you can make a system better is if you allow independent scrutiny of the system."
CBSA stated it is "actively" working to minimize bias and has "already taken several important steps" including monitoring performance across equity groups. No independent audit of the TCI has been publicly reported. CBSA's annual privacy reports for 2023–2024 and 2024–2025 record zero privacy investigations and zero privacy audits across the entire agency and make no mention of the TCI. No Algorithmic Impact Assessment for the TCI has been published on the Open Government Portal, despite CBSA having published AIAs for other systems.
The TCI operates in a consequential decision context: border entry decisions can result in secondary inspection, detention, refusal of entry, or seizure. The system scores all entrants at equipped ports, not a subset. The federal Directive on Automated Decision-Making requires completion and publication of an Algorithmic Impact Assessment and peer review by qualified experts for automated decision systems at higher impact levels.
Harms
CBSA's Traveller Compliance Indicator assigns compliance scores to travellers based on historical data that may encode past enforcement patterns shaped by officer discretion and demographic targeting. Without independent audit, the system may systematically assign higher-risk scores to travellers from specific national or ethnic backgrounds.
Evidence
2 reports
-
Expert bias concerns; CBSA TCI description; absence of independent audit
- CBSA to expand use of AI screening tool at land borders to flag higher-risk travellers Primary source
National expansion plans; timeline to all land ports by 2027
Record details
Policy Recommendationsassessed
CBSA should commission an independent algorithmic audit of the TCI before expanding beyond pilot deployment
Ebrahim Bagheri, University of Toronto (responsible AI researcher) (Jan 1, 2025)The TCI should be assessed under the federal Directive on Automated Decision-Making, with the impact assessment made publicly available
Treasury Board of Canada Secretariat (DADM framework) (Jan 1, 2023)Editorial Assessment assessed
The TCI is a population-level AI classification system operating in a high-stakes decision context — border entry — without a published Algorithmic Impact Assessment or reported independent audit. CBSA's own privacy reports make no mention of the system across two consecutive years, while the agency simultaneously plans national expansion. No AIA for the TCI appears on the Open Government Portal.
Status History
CBSA announced national expansion to all land ports by 2027; still no independent audit; annual privacy reports show zero privacy audits of the system
TCI piloted at six land ports; no independent audit conducted; expert bias concerns raised
Entities Involved
AI Systems Involved
The TCI is the AI system that assigns risk scores to border entrants
Related Records
- AI in Canadian Government Automated Decision-Makingrelated
- IRCC Machine-Learning Triage Sorts Millions of Visa Applications Using Models Trained on Historical Decisionsrelated
- AI Systems as Attack Surfacesrelated
Taxonomyassessed
Changelog
| Version | Date | Change |
|---|---|---|
| v1 | Mar 11, 2026 | Initial publication |