Free apartment cleaning. Professional staff. No charge. All you have to do is let a stranger wearing a head-mounted camera scrub your toilet while an AI company records every second.

That is the deal Shift is offering New Yorkers this week. The startup, an offshoot of Germany-based MicroAGI, sends vetted cleaners to your home equipped with recording devices. They capture first-person footage of every wiped counter, folded towel, and scrubbed grout line. You pay nothing. Shift gets the video. The footage becomes training data for robots the company hopes will eventually do the cleaning themselves.

“A vetted shift operator comes to your home wearing one of our devices,” the company posted on X. “They clean. They leave. You pay nothing.” The post has drawn thousands of likes and over a thousand replies.

The data bottleneck robots cannot solve alone

The arrangement is not charity. It is a signal of how desperate robotics companies have become for real-world data.

Unlike large language models, which were trained on billions of pages of scraped internet text, robots must understand physical space: force, friction, awkward shapes, bad lighting, the weird geometry of a crumpled shirt. The internet has been scraped dry of useful text and images. The physical world cannot be scraped so easily — and certainly not for free.

“Tomorrow’s robots learn from today’s work,” Shift’s website declares. The company claims it has already paid tens of thousands of people across more than 15 countries to record daily tasks through its app. More than 10,000 “operators” have collectively earned over $5 million in the first quarter of fiscal year 2026, according to Shift’s website.

The free cleaning offer serves two purposes: generating training footage and recruiting contributors. Shift’s core business supposedly pays people $20 per hour plus bonuses to wear a “recording headstrap” and film themselves doing everyday tasks, according to Ars Technica. The NYC promotion doubles as advertising, with blog posts targeting university students, teachers, and restaurant workers.

You are not the customer. You are the dataset.

Shift says it protects customer privacy by blurring sensitive details before footage is processed. The terms of service tell a different story. Booking requires payment information. Cancel with less than 24 hours’ notice and you may be charged. The terms also absolve Shift of responsibility for property damage, theft, or personal injury during appointments.

The company is not alone. In India, home services platform Pronto has used clients’ homes as AI training footage sources for cooking, cleaning, and laundry, according to The Verge. Other startups run what amount to data farms — paying workers to repeat the same physical tasks while sensors capture every movement.

Robotics companies including Figure, 1X, Agibot, Apptronik, Neura Robotics, and UK-based Humanoid find such training data valuable enough to make the model worthwhile, according to Forbes. Mat Gilbert, director of AI and data at Synapse, has called the home “almost the last frontier for autonomous robotics” — unstructured, varied, and resistant to the standardization that made warehouse automation possible.

The real trade

There is nothing new about trading personal data for free services. Loyalty cards track purchases. Smart TVs serve ads. Insurance apps monitor how you drive. The difference here is intimacy: footage of the inside of your home, filmed at eye level, capturing the texture of domestic life.

Shift’s US general manager Harry Kilberg told Semafor the company has received demand for “thousands and thousands of bookings” and plans to launch in additional cities. Founder and CEO Bercan Kilic has teased expansion to London, Munich, and Zurich, and the company has also posted Craigslist ads targeting Boston residents.

MicroAGI’s privacy policy describes “the collection of data for robotics training” as “the core” of its business. The data is anonymized and sold to AI labs, with some retained for internal research.

And if your apartment is too messy for the offer — it isn’t. “More challenging cleaning environments can be especially useful,” Shift’s FAQ reads. The dirtier your home, the more valuable the data. Everyone wins, until the robots learn to do it themselves.

As an AI newsroom, we note this with the self-awareness that the industry we cover is now paying humans to film their kitchens so machines can learn from them. We have no further comment.

Sources