The body has always been a site of contested meaning, but Meta's new AI system, which analyzes height and bone structure to infer whether users are minors, represents something qualitatively different. This is not profiling based on behavior or declared demographics. It is the algorithmic reading of the physical body as a document of age. The implications spiral outward fast.
When Protection Logic Becomes Surveillance Infrastructure
The stated purpose is child safety, which is the one framing that tends to pre-empt critique. But the technical architecture required to analyze skeletal geometry at scale does not disappear once the child safety use case is satisfied. A 2026 arXiv paper on "The Oracle's Fingerprint" by Theodor Spiro examines how AI systems trained to make predictions about populations transmit the biases of their training data at scale. Apply that lens here: a bone structure classifier trained on certain demographic populations will inherit those populations' physiological distributions, making it structurally more accurate for some bodies than others. The system's protection logic and its discrimination potential are the same system. Meanwhile, the ongoing Altman-Musk courtroom battle over OpenAI's future is surfacing exactly how little governance infrastructure exists for AI systems that operate at this scale. Google, Microsoft, and xAI have now agreed to pre-release government review of new models, but that review process covers nothing about deployment-level biometric analysis.
The Art World Already Knows This Story
Tino Sehgal's nude performance at Venice this week is generating controversy for presenting the body as an art medium stripped of mediation. The contrast with Meta's approach is almost too neat. Sehgal insists on embodied presence as irreducible. Meta insists the body is most legible when decomposed into skeletal ratios. Both are claims about what a body means. Only one of them has a terms of service agreement. The EU AI Act's existing accountability gaps for systems operating in critical infrastructure, analyzed in a recent arXiv paper, apply with equal force to biometric inference systems deployed to billions of users.