Data labelling: The hidden human cost behind AI's job creation

Most workers are not employees. They are classified as "users" by the platforms that distribute the work. There are no formal contracts. Workers are called on when tasks match their track record, and they have minimal rights to appeal performance assessments. User agreements exist primarily to protect the companies outsourcing the work, often barring workers from disclosing what they see, even though the datasets are already anonymised.

Workers in China face additional hurdles. They cannot access US-based crowdsourcing platforms, which tend to offer higher pay. Using a VPN to get around this risks an account ban. Chinese platforms often pay workers only after their tasks have been assessed and confirmed to meet standards, meaning hours of work can go unpaid.

The study's interviewees reported getting less work over time as AI models improve. The tasks that remain are more difficult and time-consuming. Workers expressed little concern about being replaced by AI, but the reason was not optimism. "If I don't make this money, someone else will, and I will be replaced eventually anyway," one worker said.

To retain a consistent workforce, companies have turned to recruiting more vulnerable groups. One example cited in the study is collaboration with local government initiatives supporting disabled people. These workers are less likely to quit because the job is often their last resort. Other workers cited disability, pregnancy, or recent graduation as reasons they could not find other employment.

The study's author described data workers as "disposable batteries" of the AI economy, drained of every last charge and then discarded. The finding echoes broader concerns about the quality of jobs created by the AI industry. The question is not just how many jobs are created, but what kind they are, and whether the human cost is worth the machine's gain.

Australia is accelerating its pursuit of an AI-driven economy. The study's author argues that protections and a long-term plan for this workforce are needed now, before the harm becomes embedded in the system.