Pronto's AI Training Pilot Sparks Privacy Concerns
Hızlı Bakış
- On-demand home services startup Pronto is facing backlash for a pilot program where workers wear head-mounted cameras to record household jobs for AI training.
- While the company claims data is anonymized and deleted within 48 hours, derived datasets are retained, raising privacy concerns among users and competitors.
Yapay zekâ özeti
Neden Önemli?
Pronto, an on-demand home services startup, is piloting a program where workers wear head-mounted cameras to record jobs. This initiative is intended for customers who are uneasy about having unfamiliar workers in their homes while they are away. The program is linked to training data for 'physical AI' and robotics.
BENGALURU: A customer books a cleaner through an app. A worker arrives. But this time, there is a camera strapped to the worker's head. The image has sparked widespread unease online after on-demand home services startup Pronto confirmed to TOI that a small subset of customers can opt into a programme where household jobs are recorded using visible, head-mounted cameras worn by workers. The issue gained attention after online portal Entrackr reported that an investor memo linked Pronto's workflows to "physical AI" and robotics training data. Pronto cofounder and CEO Anjali Sardana said the pilot covers 0.1% of users and is meant for customers who feel uneasy about allowing unfamiliar workers into their homes while away. "They worry about what's happening in their home during their booking. Something may be stolen or broken, or work may not be done properly." A new investor at Pronto, Lachy Groom, is also a co-founder of Physical Intelligence, a US startup building foundation models for robotics. Pronto recently raised fresh capital from him and others, doubling to a valuation of $200 million in just a few months. She said videos are anonymised, no audio is recorded, and footage is deleted within 48 hours. But the company confirmed that "derived datasets" from those recordings are retained, including "key point mapping" data tracking body joints and hand movements. When asked whether those datasets could eventually be monetised or shared with third-party AI or robotics firms, Pronto declined to comment. Customers who opt for it pay an extra Rs 29 per booking; workers are paid extra as well. Pronto's clarifications have done little to calm the backlash. On X, a post declaring "recording inside your house to train AI! This is scary" drew thousands of views and interactions. "Trust is the cornerstone of any consumer/service business and Pronto just lost it," wrote another user. Others raised a more practical concern, that homes are nearly impossible to truly anonymise. "Think name plates, ID cards, credit cards, bills," one frequent home-services app user told TOI. Rival platforms moved quickly to distance themselves. "We are in the business of trust, and we take customer privacy extremely seriously. We do not engage in any such activities, have never done so, and have no plans to do so," Urban Company cofounder Abhiraj Bhal told TOI. Snabbit founder Ayush Agarwal said his company is not undertaking any such activity either. Behind the debate is a rapidly emerging market for "egocentric," or first-person, data used to train physical AI systems. Unlike chatbots trained on text, physical AI needs exposure to real-world environments such as kitchens, utensils, shelves, clutter and repetitive human movement. Some AI labs globally pay between $4 and $10 an hour for such data, according to one founder building in the space. The legal picture remains murky. Nikhil Narendra, a partner at law firm Trilegal noted that anonymised household data could potentially fall outside India's Digital Personal Data Protection Act, while cautioning that India still lacks a governance framework for non-personal data. For now, the front door remains the last line of privacy. The debate is about how long that holds.
Bundan Sonra Ne Olabilir?
Yapay zekâ öngörüsü — kesinlik taşımaz
Further public outcry and potential regulatory investigation into Pronto's data collection practices.
Çok muhtemel · Haftalar içinde
Rival companies will continue to publicly distance themselves from Pronto's methods, emphasizing their commitment to privacy.
Muhtemel · Günler içinde
The debate around 'egocentric' data for physical AI training will intensify, leading to calls for clearer regulations.
Muhtemel · Aylar içinde
Açık Sorular
- Will Pronto monetize or share the derived datasets with third-party AI or robotics firms?
- What specific governance framework will be applied to the non-personal data collected?
- How will Pronto ensure true anonymization of household data, given the presence of personal items?
- What are the long-term implications for privacy in the home services industry?