flock, automatic The license plate reader and AI-powered camera company employs Upwork’s international employees to train its machine learning algorithms, including training materials that teach employees how to review and classify footage containing images of people and vehicles in the United States, according to documents reviewed by 404 Media that the company mistakenly published.
The findings raise questions about who exactly has access to the footage collected by Flock’s surveillance cameras and where the people reviewing the footage are based. Flock has become a popular technology in the United States, with its cameras installed in thousands of communities and used by police every day to investigate carjackings and other crimes. Local police also conducted numerous searches for ICE in their system.
Companies using AI and machine learning regularly rely on overseas workers to train their algorithms, often because the labor is cheaper than hiring domestically. But due to the nature of Flock’s business, which is building surveillance systems that constantly monitor the movements of U.S. residents, the footage may be more sensitive than other AI training jobs.
Flock’s cameras continuously scan the license plate, color, brand, and model of every vehicle it passes. This will allow law enforcement to search cameras across the country to see where else the vehicle has traveled. Authorities typically comb through this data without a warrant, leading the American Civil Liberties Union and the Electronic Frontier Foundation to recently sue a city covered in nearly 500 flock cameras.
At a high level, Flock uses AI or machine learning to automatically detect license plates, vehicles, and people (including the clothing they’re wearing) from camera footage. Flock’s patent also mentions a camera that detects “race.”
Multiple tipsters pointed 404 Media to a published online panel showing various metrics related to Flock’s AI training.
This includes numbers for “completed annotations” and “annotator tasks remaining in the queue,” where annotations are notes that workers add to the reviewed footage to help train the AI algorithm. Tasks include classifying vehicle make, color, and type, transcribing license plates, and “voice tasks.” Flock recently started promoting its ability to detect “screams.” The panel showed that workers may create thousands of annotations in two days.
The panel released included a list of people tasked with annotating Flock’s footage. Based on these names, 404 Media found that some of them are located in the Philippines, according to their LinkedIn and other online profiles.
Many of these people were hired through Upwork, according to the documents exposed. Upwork is a gig and freelance work platform that allows companies to hire designers and writers and pay for “AI services,” according to its website.
Tipsters also pointed to several publicly available Flock presentations that explained in more detail how workers sorted the footage. It’s unclear exactly what kind of camera footage Flock’s AI employees review. However, screenshots included in the worker guide show numerous images taken from vehicles with U.S. plates, including in New York, Michigan, Florida, New Jersey and California. Other images include road signs that clearly indicate that the footage was taken in the United States, and one image includes an advertisement for a particular law firm in Atlanta.
