Is Your AI Being Trained by Workers Under Surveillance?
European privacy complaints against Scale AI's Outlier platform allege extensive worker monitoring. Why buyers should know who trains their models and under what conditions.
Short answer
noyb's complaints against Scale AI's Outlier platform are allegations, not proven breaches — but they ask whether responsible AI use is credible when training workers face intrusive monitoring. Know subcontractors, contract for proportionate privacy standards, obtain evidence beyond promises, and protect the quality of human review.
When we use artificial intelligence, we rarely think about the people who helped train it. Behind the clever answers and impressive capabilities are thousands of workers checking responses, correcting mistakes, and teaching systems what good answers should look like.
But who is looking after those workers?
On 27 October 2026, European privacy organisation noyb announced five complaints against Scale AI, the American company operating the Outlier platform. Outlier connects more than 900,000 freelance workers with assignments used to train and improve artificial intelligence. Its reported customers include some of the world's largest technology companies, including Meta, OpenAI, and Google.
The complaints allege that workers were subjected to extensive digital surveillance, including monitoring of their online activity, keyboard use, and mouse movements. Researchers identified up to 19 tracking systems operating while people worked. They also reported that workers were repeatedly required to provide identity documents and facial videos through a separate identity-verification service. Some workers allegedly received no response when requesting copies of their personal information.
These are allegations, not established breaches of European privacy law. The relevant authorities have yet to determine whether the company's practices were unlawful. Nevertheless, the complaints raise an uncomfortable question: can a company claim to use responsible AI when it does not know how the people training that AI are being treated?
Why this matters
Most businesses purchase AI services without seeing the human effort behind them. A company might carefully protect its own customers' privacy while unknowingly relying on suppliers whose workers face intrusive monitoring or unclear data-handling practices.
There is also a practical concern. Workers who feel constantly watched, rushed, or threatened with losing assignments may be less willing to challenge questionable results. That matters when their job is to identify mistakes and unsafe behaviour in an AI system.
What to do
Know your suppliers. Identify who trains, checks, and maintains your AI systems, including subcontractors. Ask what information is collected. Establish whether workers are monitored, why monitoring is necessary, and how their information is protected.
Check the contracts. Require reasonable privacy standards, clear responsibilities, and the ability to investigate serious concerns. Look beyond promises. Obtain evidence that suppliers actually follow their stated policies.
Protect the quality of human review. Make sure workers can report mistakes and unsafe outcomes without inappropriate pressure. The objective is not to prohibit monitoring altogether. It is to ensure that any monitoring is justified, proportionate, and properly controlled.
Sources
- Reuters, 27 October 2026 — privacy group complaints against Scale AI.
- noyb, 27 October 2026 — complaint regarding Outlier (title reflects noyb's allegation, not a regulatory finding).
- ISO/IEC 42001 explained — International Organization for Standardization.
We map who is involved in your AI supply chain, what information is collected, and where responsibilities begin and end — through gap assessment and supplier oversight under ISO/IEC 42001. Responsible AI is not just what a machine produces; it is also how the machine was made.