At any given moment, thousands of third-party contractors worldwide are engaged in reviewing and labeling content on Facebook and Instagram. This process is crucial for training artificial intelligence (AI) and enhancing new product features. However, because these contractors can access both public and private user posts, many see this as a potential violation of privacy.
According to Reuters, approximately 260 contract workers in Hyderabad, India, have dedicated over a year to labeling millions of Facebook posts dating back to 2014. Their tasks include identifying the subject matter, occasion, and intent of the posts. Facebook has stated that this information is used to develop new features and potentially boost user engagement and advertising revenue.
Globally, Facebook oversees around 200 similar content labeling initiatives, many of which are designed to improve its AI capabilities. While such "data annotation" practices are common in the tech industry—helping AI to recognize elements like traffic lights or pedestrians in videos—this may not alleviate users' privacy concerns.
Contractors in Hyderabad report encountering a wide range of content, from text updates to photos, videos, and Stories, including those shared privately. Despite Facebook's commitment to a "private future" platform, one employee remarked to Reuters that these practices are unlikely to change, as they remain integral to AI training and product development.