Apple trained a foundation model on 2.5 billion hours of wearable data from 162,000 people that can infer age within ~2.5–4 years, identify sex with near‑perfect accuracy, detect pregnancy, and flag infection weeks. This shows passive behavioral signals can reliably reveal sensitive health states without explicit tests. The capability leap raises questions about consent, secondary use, and who controls inference rights—not just data collection.
— If consumer wearables enable medical‑grade inferences, regulators must address privacy, liability, and data‑rights frameworks before insurers, employers, or platforms weaponize these predictions.
EditorDavid
2025.09.22
90% relevant
Apple trained an algorithm on large wearable datasets and won FDA approval to alert users of possible high blood pressure—an example of consumer wearables making medical‑grade inferences (age, sex, pregnancy in prior work; now hypertension risk) with major privacy and governance implications.
BeauHD
2025.09.09
80% relevant
Apple is adding hypertension notifications and sleep-scoring trained on large datasets (100,000+ people; 5 million nights) and seeking FDA clearance, reinforcing the trend that consumer wearables can infer sensitive health states at scale and raising privacy/rights questions.
Alexander Kruel
2025.08.24
100% relevant
Apple’s arXiv paper reporting the 2.5B‑hour wearable foundation model and its age/sex/pregnancy/infection inference performance.
← Back to All Ideas