MicroAGI offers free home cleaning in New York

A German startup called MicroAGI is offering New York City residents two-hour home cleanings at no charge, on one condition: the cleaners wear cameras and record the work to generate training data for robotics. The program runs through a newly launched app called Shift, which the company began promoting on social media in late May, according to Ars Technica.

Shift’s pitch is that “trusted professional house cleaners” will visit apartments and homes while capturing first-person video of routine tasks—wiping counters, picking up clutter, handling tools—footage MicroAGI says it uses to accelerate “embodied AI”. To book, customers must provide contact details, a home address and access instructions, and the app also requires payment information; the terms allow charges for late cancellations or if the customer is not available to let the cleaner in, Ars Technica reports.

The offer illustrates how valuable mundane, high-resolution behavioural data has become in the race to build robots that can operate in unstructured environments. Industrial robots learn in controlled spaces; homes are the opposite, full of brand-specific objects, irregular layouts and improvisation. Hiring humans to generate “ground truth” video turns domestic life into a data factory, and the cost structure helps explain why startups reach for consumer-facing schemes rather than waiting for robots to collect the data themselves.

MicroAGI says names and faces are anonymised and sensitive details blurred before footage is used, with “irreversible transformations” applied before uploads to cloud servers. But the company’s own materials leave open what happens when the background is the identifying feature: distinctive interiors, valuables, documents on a table, or the simple fact that an address and access instructions have already been handed over to a platform that disclaims responsibility for property damage, theft, or personal injury during appointments.

Ars Technica notes that Shift’s site does not clearly state whether customers can demand deletion of their home-cleaning videos from training datasets. That omission matters because, once data is incorporated into model training and shared internally or with partners, removal becomes a policy question rather than a technical one. The arrangement also shifts risk in a familiar direction: the startup keeps the asset that appreciates (the dataset) while the customer supplies the setting and bears much of the downside if anything goes wrong.

MicroAGI’s US general manager, Harry Kilberg, told Ars Technica the platform already pays tens of thousands of people across 15 countries to record daily work and chores, and the company markets payments for contributors who capture video. In New York, the bargain is simpler: two hours of labour now, and a permanent training example later.

The service is advertised as free cleaning in exchange for footage. The terms of service, Ars Technica reports, explicitly limit the platform’s liability if the visit does not go as planned.