Google says it used field research that offered people $5 to scan their faces, to improve accuracy of face unlock; research data will be kept for 18 months
It's ensuring that the Pixel 4 will work on a wide range of faces — Google just announced a face unlock feature for the upcoming Pixel 4 …
Context & Ripple Effects
This is the origin document for the Pixel 4's biometric pipeline: alongside the face unlock and radar-powered Motion Sense announcement, Google discloses it paid participants $5 each for facial scans to broaden the system's accuracy across face types, and that it will retain that research data for 18 months.
The disclosure matters because everything that followed in the coverage traces back to how this data was gathered and what the model missed — the contractor Randstad's allegedly dubious recruitment of people of color for scans, and the confirmed finding that Face Unlock worked on users with closed eyes.
First-order effects
- Participants who sold face scans for $5 have their biometric data held by Google for 18 months, a retention window set before any of the later controversies forced policy scrutiny.
- Pixel 4 buyers get a face unlock trained on this dataset, aimed at working across a wide range of faces at launch.
Second-order effects
- The collection method becomes Google's liability: reports about Randstad's targeting practices put the $5-scan program's consent standards under pressure just as the phone ships.
- The closed-eye vulnerability forces Google into a remediation cycle — first promising an optional open-eyes setting within weeks of launch, then shipping it as the default behavior seven months after release.
Third-order effects
- If the pattern holds, biometric authentication features ship on aggressive timelines and get hardened retroactively, making post-launch security patches a standing part of face-unlock product cycles rather than an exception.
- How vendors pay for and retain training biometrics — consent, compensation, deletion windows like the 18-month promise here — moves toward being a compliance question regulators and plaintiffs treat as inseparable from the feature itself.
The trend: Smartphone makers are racing facial-recognition features to market on crowdsourced scan datasets faster than they can enforce consent and security standards, leaving defaults to be corrected after launch.