Herd, automatic license plate reader and artificial intelligence camera company is using overseas Upwork workers to train machine learning algorithms, and training materials tell workers how to review and categorize footage, including photos of people and vehicles in the United States, according to material reviewed by 404 Media that was accidentally leaked by the company.
The findings raise questions about who exactly has access to the footage collected by Flock’s surveillance cameras and where those viewing the footage might be located. Flock technology has become ubiquitous in the U.S., with cameras in thousands of communities that police officers exploit every day to investigate cases such as car thefts. Local police too conducted numerous searches for ICE in the system.
Companies using artificial intelligence or machine learning regularly turn to foreign workers to train their algorithms, often because labor is cheaper than hiring domestically. However, the nature of Flock’s business – creating a surveillance system that constantly monitors the movements of US residents – means the recordings may be more sensitive than other AI training tasks.
Flock’s cameras continuously scan the license plate, color, make and model of all passing vehicles. Law enforcement can then search cameras across the country to see where else the vehicle has traveled. Authorities typically review this data without a warrant, which prompted the American Civil Liberties Union and the Electronic Frontier Foundation recently sued the city captured in almost 500 Flock cameras.
Generally, Flock uses artificial intelligence or machine learning to automatically detect license plates, vehicles, and peopleincluding what clothes they are wearing based on camera footage. Patent for the herd also mentions cameras that detect “race”.
Many tipsters pointed 404 Media to an exposed web dashboard that showed various metrics related to Flock’s AI training.
It included data on “annotations completed” and “annotator tasks remaining in the queue,” where annotations are notes that workers add to the footage they review to support train artificial intelligence algorithms. Tasks include categorizing vehicle makes, colors and types, transcribing license plates and “audio tasks”. Flock recently started advertising a feature which will detect the “scream”. The panel showed that workers sometimes created thousands of annotations in two days.
The unveiled panel included a list of people tasked with describing Flock’s footage. Given these names, 404 Media said that according to their profiles on LinkedIn and other websites, some of them were located in the Philippines.
The leaked materials show that many of these people were employed by Upwork. According to the Upwork website, Upwork is a work platform for freelance artists where companies can hire designers and writers or pay for “artificial intelligence services.”
Tipsters also pointed to several publicly available Flock presentations that explained in more detail how employees were supposed to categorize the recordings. It’s unclear what specific camera footage Flock’s AI employees are reviewing. However, screenshots included in employee handbooks show numerous photos of vehicles on U.S. plates, including those in Fresh York, Michigan, Florida, Fresh Jersey and California. Other photos show street signs clearly showing that the footage was shot in the US, and one photo shows an advertisement for a specific Atlanta law firm.
