AI-Assisted Facial Recognition for Passport Identity Verification

Ask about this system

Questions go to the Helpful Places, not the vendor.

Your rights

• Right to Be Informed of AI UseApplicants are entitled to be informed that facial recognition technology is used during the passport application process. The AIA requires that a plain language notice be posted through all service delivery channels in use (Internet, in person, mail, or telephone). This AIA is publicly available through the Government of Canada Open Data portal.
• Right to an Explanation of a DecisionApplicants who receive a denial of service or are subject to regulatory action are entitled to a meaningful plain language explanation that includes: the role of the facial recognition system in the decision; the training and client data used; the criteria applied to evaluate their data; the output produced by the system; and a justification of the administrative decision including the principal factors that led to it. Information about recourse options must also be provided where appropriate.
• Right to a Human ReviewAll cases where the facial recognition system fails a comparison are referred to human officers for assessment and decision. A multi-stage escalating review process is in place. All decisions to refuse to issue a travel document, withhold service, or revoke a travel document are made by a human and never based solely on facial recognition results. The system supports human override and logs all override instances.
• Right to ContestA recourse process is established for clients who wish to challenge a decision related to their passport or travel document application. The AIA confirms that a recourse process exists and that clients will be informed of relevant recourse options alongside any meaningful explanation of a decision resulting in denial or regulatory action. Contact information for recourse is not specified in the AIA document.
• Right to Non-discriminationApplicants have the right to non-discriminatory treatment. The program acknowledges that facial recognition performance can vary by gender, age, and skin tone. Mitigations include requiring high-quality standardized photos, using an algorithm tested for performance across diverse populations by NIST, and ensuring all adverse decisions are made by a human officer rather than solely by the system. A Gender-based Analysis Plus (GBA+) has been conducted and bias-testing processes are documented.

Risks and safeguards

• Civil liberties harmThe use of facial biometrics may engage a client's reasonable expectation of privacy; performance variability across gender, age, and skin tone may produce higher false match rates for certain demographics.Safeguard: A four-part privacy necessity test was completed per the Office of the Privacy Commissioner's recommendation. Privacy Impact Assessments were completed in 2003 and 2011; a new PIA is underway for 2024-25. Biometric data is classified Protected B/C. Access is restricted to IRCC Passport Program staff. The algorithm is tested against the NIST framework. False matches are adjudicated by human officers and do not affect service standards. All final decisions to refuse or revoke travel documents are made by a human.
• Reputational harmFalse positive matches could incorrectly associate an applicant with a fraudulent or flagged identity, causing reputational harm.Safeguard: A multi-stage human review process is in place for all failed comparisons. The human reviewer at the first stage is deliberately not shown the system's similarity score to avoid anchoring bias, and does not know the type of comparison that generated the case. Cases escalated through multiple human review stages before any adverse action is taken. All decisions to refuse, withhold, or revoke travel documents are made by humans and never solely on the basis of facial recognition results.