Pronto’s Physical AI initiative and the privacy debate

Bengaluru-based household service startup Pronto has initiated a pilot program capturing video footage of its staff performing tasks like cleaning, folding, washing, and meal preparation at users’ homes. The data will be used to train physical AI systems and robots, according to an Entrackr report citing investor documents from Glade Brook Capital. The memo states Pronto is formalizing India’s informal labor markets while generating real-world training data for physical AI labs, with promising initial partner interest. Pronto confirmed the pilot involves AI-driven data initiatives, emphasizing customers can voluntarily consent to recording service sessions. The company aims to commercialize the initiative, leveraging its workforce to collect data for robotics research. Physical AI differs from traditional AI by interacting with the real world through machines, robots, or devices equipped with computer vision, sensor processing, and machine learning. Unlike digital AI trained on text or images, physical AI requires real-world environments—like a user’s home—to observe and record actions, such as hand movements, for simulation-based training. IBM explains that robots need simulated environments replicating real-world conditions, including lighting, friction, and object interactions. Pronto’s recorded data will feed into these simulations, where robots repeatedly attempt tasks until they achieve a success threshold. Once refined, the robots transition to real-world applications. The move has raised privacy debates among users, investors, and competitors, with founders in India Inc. weighing in on the ethical implications. Entrackr’s report highlighted the internal memo’s focus on formalizing labor markets through data collection, marking a shift toward physical AI training in household settings.