- A Roomba robot vacuum captured intimate images, including a woman on the toilet, during development testing in 2020.
- These images were sent to Scale AI for data annotation and later leaked by gig workers in Venezuela on social media.
- iRobot confirmed the images came from modified development devices, not consumer products, with consent from paid collectors.
- Data annotation involves human workers globally labeling images for AI training, creating privacy risks in the supply chain.
- Robot vacuums with cameras collect invasive data; companies use real-home images to train algorithms for object recognition.
- Privacy policies often allow broad data use for product improvement, but consumers may not understand the extent of sharing.
- The incident highlights gaps in data protection, as facial images, including minors, were not consistently obscured.
- The growing data annotation market relies on low-paid contract workers, making sensitive data vulnerable to leaks.
- Legal frameworks in the U.S. lack comprehensive privacy laws, relying on company policies that may not fully protect consumers.
- Robot vacuum companies aim for broader home robotics, increasing data collection needs and privacy concerns.