Federated learning trains AI models by sending algorithms to your device instead of sending your data to a central server, leaving sensitive information on your phone while only sending back refined patterns to improve the model. The privacy gains are real but incomplete—the model updates you send back can still leak information about your data, and servers still see aggregate patterns that can reveal population-level secrets.
Federated learning is a machine learning approach where AI models are trained directly on your device using your local data, with only aggregated model updates — not raw personal data — sent back to central servers. Tech companies promote this as a privacy-preserving alternative to uploading everything to the cloud, but the technique introduces its own set of risks and limitations that users rarely understand.
Despite keeping raw data local, federated learning updates can still leak sensitive information through gradient inversion attacks, and the models trained on your device may still inform behavioral targeting systems. Understanding the genuine privacy trade-offs of on-device AI helps you critically evaluate product claims and decide which features are worth enabling based on your personal threat model.
Peri can explain this concept, give practical examples, help you decide whether it applies to your situation, or recommend a journey if appropriate.
Explore related journeys or tell Peri what you're working through.