A look at how Amazon Rekognition can be used to build facial recognition tools at incredibly low cost by anyone with a computer
Thomas Fox-Brewster / Forbes :
Context & Ripple Effects
AWS spent late 2017 steadily widening Rekognition's surface area — a real-time face search update in November, then Rekognition Video weeks later for analyzing live video streams. The Forbes piece is the logical endpoint of that roadmap: Thomas Fox-Brewster demonstrates that the API's pay-as-you-go pricing means facial recognition is no longer a capability you need a lab or a budget to deploy — a computer and an account suffice.
That framing matters because it reframes Rekognition from an enterprise product into a commodity input, and it anticipates the fight that follows in the corpus: Andy Jassy's staff-meeting defense arguing regulators, not Amazon, should set the rules, and subsequent reporting that Rekognition's only known law enforcement client skips the confidence threshold Amazon cites against bias findings.
First-order effects
- Surveillance-grade face matching becomes accessible to any individual or small organization with a computer, removing the cost barrier that previously limited such tooling to well-funded institutions.
- Amazon faces immediate reputational exposure: the same low-cost accessibility the demo celebrates is what puts its misclassification rates — 19% for women overall, 31% for darker-skinned women per the testing cited — under direct comparison with IBM and Microsoft.
Second-order effects
- Competing cloud vendors are pushed to differentiate on accuracy across demographics rather than price, since the relationships data shows Rekognition already trails IBM and Microsoft on those error rates.
- Civil-society and press scrutiny shifts from whether governments buy the tool to who else can — the Forbes demo effectively widens the set of actors regulators and Amazon must account for beyond its known law-enforcement client base.
Third-order effects
- If cloud APIs keep commoditizing biometric capabilities, the burden of governing facial recognition lands almost entirely on legislation rather than vendor policy — exactly the position Jassy articulates when he calls regulation the government's duty.
- The pattern points toward accuracy and threshold settings becoming the central accountability battleground, since the corpus shows even the sole known police client not using the confidence threshold Amazon leans on in its defense.
The trend: Cloud AI APIs are turning once-exclusive surveillance capabilities into cheap commodities faster than governments are writing the rules to govern them.