How Pronto’s Physical AI Pilot Raised Questions Around Consent & Surveillance

What happens when everyday household chores stop being just labour and start becoming training data for AI systems? And what does it mean for consumers if activities inside their homes are quietly helping train robots?

These questions are now at the centre of a growing debate around India’s booming instant home services market, after quick service startup Pronto came under scrutiny for recording videos inside customers’ homes as part of a pilot linked to physical AI and robotics systems.

At the centre of the controversy is a possibility many consumers had not thought about before — everyday household activities such as washing dishes, folding clothes, cleaning kitchens and organising homes could become valuable training data for AI-powered robots.

The controversy first surfaced after Entrackr reported that Pronto was exploring ways to generate “real-world training data” for physical AI and robotics systems using footage captured during household chores.

Investor documents reviewed by the publication stated that Pronto was “developing a data business leveraging its workforce to capture real-world household data for robotics labs” and was already “piloting real world training data with leading physical AI labs.” The report said Pronto’s service professionals were wearing outward-facing cameras while carrying out tasks such as washing dishes, folding laundry, cleaning homes and meal preparation.

Pronto later confirmed the pilot, but stressed that it was “strictly opt-in,” enabled only when customers specifically chose the feature during booking. Further, the startup mentioned that the feature is currently limited to less than 0.01% of users. 

“Pronto does not, by default, record in customers’ homes, nor does it use customer data. We started the pilot you are referring to a few days ago. It is a strictly opt-in feature, available only to customers who actively choose it at the time of booking. The customers who have enrolled in the pilot programme pay for the feature and they receive anonymised footage in the app for 48 hours. There are specific use cases where customers prefer to have a recording, such as wanting a job done at home while they are away and having the reassurance of seeing afterward that the work was done the way they expected,” Pronto told Inc42.

In a social media post, the startup added that it spent months ensuring compliance with India’s Digital Personal Data Protection (DPDP) Act and argued that physical AI systems require real-world behavioural data to train robots to function inside household environments. 

“And we are not the only company in the space doing this,” Pronto’s post read.

According to Pronto, first-person videos of repetitive household tasks performed in real homes could become a “foundational data layer” for physical AI.

Further, the startup said it is not sharing data with any third-party AI labs. “And the video itself will never be shared with anyone and is deleted from our servers in 48 hours,” it claimed.

However, critics and social media users quickly questioned whether the startup’s stated deletion policy aligned with the broader goal of building AI training datasets, since such data typically requires annotation, curation, storage and processing before becoming commercially useful.

The backlash triggered wider concerns around privacy, consent and transparency. Many argued that users may not fully understand how recordings made inside their homes could eventually feed AI systems — potentially normalising household surveillance and turning private spaces into data sources for physical AI companies.

The controversy grew loud enough to draw the central government’s attention. According to a Moneycontrol report, MeitY has taken cognisance of the matter and is examining concerns around surveillance, consent, and the use of customer-home data for AI systems.

Urban Company, Snabbit Publicly Distance Themselves

A day after this, Snabbit came under scrutiny as another report by Entrackr said that the startup was previously approached by Human Archive — a startup building datasets for physical AI and robotics systems. 

Snabbit told Inc42 that it had explored the technology and conducted a limited assessment in a controlled training-centre environment. However, it refrained from rolling the tech out inside customer homes.

“The limited assessment was a one-time evaluation conducted within a controlled training-centre environment to study the proposal and understand whether such technology could have any relevance in a training-centre setting. Following the assessment, Snabbit did not move forward with the proposal, including at a training-centre level,” the startup told Inc42. 

However, Snabbit informed that it had entered a standard non-disclosure agreement (NDA) to facilitate preliminary exploratory discussions and confidential evaluation of a proposal shared by Human Archivy. “To be absolutely clear, the NDA did not create or imply any form of commercial or operational relationship between the companies,” it added.

The heat was also faced by Urban Company, which runs instant househelp platform InstaHelp. Its founder Abhiraj Singh Bhal also took to social media to clarify the company’s stance on the issue. In a post on X, Bhal said that it neither records inside customers’ homes nor plans to do so in the future.

“We are in the business of trust, and we take customer trust and privacy extremely seriously. We do not engage in any such activities, have never done so in the past, and have no plans to do so in the future,” he said.

Human Archive’s Role And The Bigger Physical AI Push

As the controversy deepened, one name began looming large over the debate — Human Archive. Questions soon emerged over what role the startup played and why multiple home services startups were being linked to it. 

Human Archives, backed by Y Combinator, is focused on building large-scale multimodal datasets for embodied AI and humanoid robotics systems.

Founded in 2026 by Raj Patel, Shloke Patel, Rushil Agarwal and Samay Maini, the startup develops specialised hardware systems fitted with cameras and sensors to capture detailed recordings of human activity in real-world settings. 

These systems are deployed across homes and industrial spaces, generating structured datasets that help robots better perceive, understand and interact with their surroundings.

The startup’s goal is to accelerate the development of autonomous machines capable of performing complex physical tasks. The datasets can power a range of use cases, from industrial automation and warehouse robotics to household assistants and service robots.

Cofounder Raj Patel said that the startup had approached multiple home services platforms, including Urban Company, before eventually working with a smaller player.

Patel claimed users were offered subsidised services in exchange for recording consent and claimed that nearly 98% of users agreed to be recorded when given a cheaper option. He further claimed the experiment scaled from 48 to 1,315 bookings per day within two weeks across multiple Indian cities, including Delhi, Gurugram, Mumbai, Jaipur, Indore, Guwahati, and Patna.

He also suggested that wearable cameras, gloves, and full-body motion capture systems could eventually become standard across labour networks linked to physical AI training.

The comments sparked alarm because they suggested India’s low-cost labour ecosystem and home services market could evolve into a large-scale data pipeline for global robotics companies.

The AI Gold Rush No One Consented To

The controversy did not emerge in isolation. Just weeks earlier, in April, a viral video offered an unsettling preview of where this trend was heading.

The footage showed Indian factory workers sewing garments while wearing head-mounted cameras fitted to white headbands, recording their hand movements throughout their shifts. The video raised the question of whether factory workers are unknowingly training AI to ultimately replace them.

An investigation by Scroll.in identified factories where workers were being asked to wear the cameras during work hours so companies could analyse how tasks were performed on the factory floor.

Many said consent was unclear or not properly obtained. Others reported physical discomfort from wearing the devices for long hours, including heat and irritation, alongside broader concerns about surveillance.

The companies behind such data collection describe the footage as egocentric video, captured from a first-person perspective and processed into datasets used to train AI and robotics systems. They collect large-scale labour-sourced video data from factory workers and package it into datasets for global AI firms working on robotics and computer vision.

That is what makes the Pronto debate feel much bigger than one startup. The deeper concern is whether platforms operating inside homes and factories are quietly evolving into AI data infrastructure, and whether workers and consumers truly understand what they are consenting to.

Investor interest reflects a broader global shift toward robotics and embodied AI. Many investors now believe the next generation of AI winners will be built on proprietary real-world behavioural datasets, not just computing power and models.

The controversy raises difficult questions — What counts as informed consent under India’s Digital Personal Data Protection Act? Who owns the data generated inside homes and workplaces? And who is accountable if recordings are leaked or misused?

As India’s home services market expands, the line between a convenience app and an AI data company may be harder to distinguish than most consumers realise.

The post How Pronto’s Physical AI Pilot Raised Questions Around Consent & Surveillance appeared first on Inc42 Media.