Complaint from data protection NGOs: Monitored for AI training

Data protection associations from several European countries have filed a complaint against the operator of a large platform that hires human workers to train AI. The company employs workers on a self-employed basis and exposes them to invasive surveillance, criticizes the Austrian NGO Noyb, which is involved in the complaint.

When training artificial intelligence, the manufacturers of common AI chatbots such as ChatGPT, Claude or Gemini also rely on human workers. Among other things, their job is to check what the AI ​​throws out - and to provide feedback accordingly. For example, they should evaluate whether the bot's answer matched the prompt. Not only inappropriate answers, but also content that contradicts the respective provider's rules should be removed from the system if possible.

This work is often undertaken by workers who are formally offered the jobs on a freelance basis via a platform. In this specific case, the complaints are directed against the industry giant Outlier, which belongs to the US group Scale AI. According to Noyb, the platform employs around 900,000 freelancers. Noyb not only criticizes the basic business model of formal self-employment, which is widespread in the platform economy and which makes workers quickly interchangeable. In this case, she also criticizes the specific conditions.

Outlier exposes its employees to “invasive tracking”. The platform places trackers on the devices of those affected, which send personal data about workers to third parties every few seconds, such as contact details and IP addresses. In addition, the platform requires workers to provide biometric identification with video recordings of their faces and upload their own ID. There is no legal basis for the use of trackers. Both the trackers and biometric identification violated the data minimization principle set out in the General Data Protection Regulation.

“Such comprehensive tracking during work really contradicts all reasonable expectations,” criticizes Lisa Steinfeld Leiser, data protection lawyer at Noyb. Baptiste Delmas from the French organization Intérêt à Agir says: “The violations highlighted make it clear: the AI ​​boom relies on a huge number of workers whose privacy is at stake.” Now it is up to the data protection authorities to take action against it.

In a fundamental statement on its website dated late last year, Scale AI emphasizes that it is making efforts to combat “modern slavery” and human exploitation. It also states that the company attaches great importance to “adhering to the highest standards of ethical business conduct because we set incredibly high standards for ourselves.” Scale AI did not respond to a query from the taz about the data protection associations' complaints.

The next step is now the turn of the supervisory authorities to whom the NGOs have submitted the complaints. They initiate proceedings and examine the allegations. It will probably take at least several months until there is a result.