- AI gig workers raise concerns over data use and working conditions
- Tasks reportedly included social media scraping and sensitive content
- Industry faces growing questions over ethics and transparency
The rapid rise of artificial intelligence has created a vast, largely unseen workforce tasked with training the systems that power it. But new accounts from workers linked to Scale AI suggest the reality of that work may be more complicated than advertised.
Scale AI, partly owned by Meta, runs a platform called Outlier, where contractors are recruited to help refine AI models. The work is often pitched as flexible, skilled employment for people with backgrounds in fields such as medicine, economics and science.
However, several workers say the tasks extended well beyond technical training. Instead, they described assignments involving scraping social media platforms like Facebook and Instagram, tagging individuals and analysing personal content.
One contractor reportedly said users would be surprised to know how their data was being handled. “I don’t think people understood quite that there’d be somebody… looking at your profile, using it to generate AI data,” they said, as quoted in a news report.
From expert tasks to uncomfortable assignments
Workers who spoke about their experience said the nature of assignments could vary widely, and at times become difficult to handle.
Some described being asked to transcribe explicit audio content or label distressing images. Others said they encountered material they had not expected, including violent scenarios or sensitive imagery.
One doctoral student said they had been assured certain content would not appear, but later encountered otherwise. “We had already been told… no nudity… no gore,” they reportedly said. “But then I would get an audio transcript thing for porn,” as quoted in a news report.
Beyond content concerns, there were also questions about how data was sourced. Several workers said they were asked to analyse publicly available social media profiles, including identifying people, locations and relationships. Some assignments, they claimed, appeared to involve data from younger users as well.
A source familiar with the company’s operations said tasks do not involve private accounts and that contributors are not required to engage with material they find uncomfortable.
A growing industry with blurred lines
The scale of this kind of work is expanding quickly. Glenn Danas, a lawyer representing AI gig workers, estimates that hundreds of thousands of people globally are now involved in similar roles across platforms.
Many workers said they took on the work as a way to supplement income, particularly in a labour market where AI itself is beginning to reshape job opportunities.
At the same time, concerns about pay and conditions have emerged. Some workers described inconsistent earnings and a project-based system where tasks could disappear without notice. Others alleged recruitment practices that promised higher pay than what was ultimately offered, though the company has disputed this.
There were also reports of monitoring software being used during tasks. Workers said tools could track activity and capture screenshots, while a source said such systems are intended to ensure accurate payment rather than continuous surveillance.
Scale AI has said its platform offers flexible, project-based work and that contributors can choose when and how they participate.
Questions over data, ownership and the future
The situation highlights a broader issue facing the AI industry. As models grow more advanced, the demand for large volumes of labelled data continues to increase. That data often comes from real-world sources, raising questions about consent, ownership and usage.
Scale AI has worked with major technology firms including Google and OpenAI, as well as government clients. Some workers said they believed their tasks were contributing to training systems used across the industry.
At the same time, many expressed uncertainty about what exactly they were helping to build. One worker reportedly questioned why certain tasks were necessary, noting that some appeared repetitive or unclear in purpose.
Despite these concerns, most said they continue to take on assignments. For many, the work remains one of the few accessible options in a changing job market shaped increasingly by automation.
“I have to be positive about AI because the alternative is not great,” one worker reportedly said.
The debate around AI is often framed in terms of future potential. But for those already working behind the scenes, the questions appear more immediate — about how data is used, how workers are treated, and where the boundaries should be drawn.
