Retail Facial Recognition Should Never Be Treated as a Casual Default

High Risk Topic

In Australian retail, facial recognition is not just another analytics feature. It is a biometric surveillance issue with major privacy implications, especially when used to identify people against a database.

The first distinction that matters is the difference between facial verification and facial identification. Verification is a one-to-one match against a known credential or stored template. Identification is one-to-many matching against a broader database or watchlist. In a retail setting, that second category is significantly more intrusive and should never be treated as a normal plug-in upgrade just because the technology exists.

The OAIC’s retail facial recognition guidance makes clear that biometric templates and facial recognition data are sensitive information. Retailers need to think about necessity, proportionality, alternatives, notice, security, and whether the claimed benefit genuinely outweighs the privacy intrusion. Recent OAIC decisions involving Bunnings and Kmart should make retailers very cautious about assuming customer watchlist systems are easy to justify.

Where Face Technology May Be Easier to Reason About

A more limited one-to-one access-control use case for staff-only secure areas is easier to reason about than one-to-many customer identification. Even then, it is still handling biometric information and still needs careful privacy, security, notice, and governance treatment. If the business simply wants staff entry control, it may still be worth asking whether cards, PINs, or other less intrusive methods would achieve the same goal. Where the site does want entry control technologies, the practical product conversation usually sits under access control rather than general retail floor CCTV.

What Retailers Should Do Before Going Further

• Define the exact operational purpose.
• Separate one-to-one verification from one-to-many identification.
• Assess whether less intrusive alternatives are available.
• Think about notice, consent issues, retention, security, and access.
• Do not assume that because a vendor can supply the technology, the business should use it.

If the retailer’s real goal is better incident review rather than biometric matching, the stronger path is usually ordinary CCTV done properly: define retention with the CCTV Storage Calculator, mark entrance and store-floor views in the Camera Planner, use the CCTV Signage Generator wherever monitored-area notice is required, and run the proposed deployment through the CCTV Compliance Checker if the business wants an extra compliance-oriented review step.

Frequently Asked Questions

• Is facial recognition in Australian retail a high-privacy-risk technology?

  Yes. Facial recognition in retail settings is a high-privacy-risk technology because it involves biometric information and can be highly intrusive, especially when used for one-to-many identification.

• What is the difference between facial verification and facial identification?

  Facial verification is one-to-one matching against a known template, while facial identification is one-to-many matching against a larger database. The second model is generally much more intrusive in a retail setting.

• Should retailers treat facial recognition as a normal default security upgrade?

  No. Retailers should not treat facial recognition as a casual default. They should first assess privacy risk, proportionality, legal obligations, and whether less intrusive alternatives would achieve the same purpose.

• Where might face technology be more defensible than customer watchlisting?

  A more limited one-to-one access control use case for staff-only secure areas is easier to reason about than broad one-to-many customer surveillance, but it still requires careful privacy and security governance.

• Is one-to-one verification different from one-to-many identification?

  Yes. One-to-one verification checks whether an enrolled user matches their own stored credential, while one-to-many identification tries to identify a person from a wider database. The privacy implications are very different.

• Should face recognition be the only credential on a site?

  Not always. Many sites still want a fallback such as cards, PINs, or app credentials so operations can continue if a user or environment does not suit face-only entry.

Sources and Further Reading

• OAIC: Facial Recognition Technology – A Guide to Assessing the Privacy Risks
• OAIC: Bunnings Breached Australians’ Privacy With Facial Recognition Tool
• OAIC: Kmart’s Use of Facial Recognition to Tackle Refund Fraud Unlawful
• OAIC: Facial Recognition Technology in Retail Settings After the Bunnings Decision