The conversation about AI privacy usually focuses on the people whose data gets used to train AI models. Less attention goes to the people doing the actual labeling, rating, and annotation work — the contractors hired by platforms like Scale AI’s Outlier, Surge AI, and Remotasks.
A complaint filed by the European privacy group noyb against Scale AI’s Outlier platform in late 2026 changed that. The complaint alleged unlawful tracking of freelance workers: persistent behavioral profiling, covert data collection, and consent mechanisms that arguably didn’t meet the bar set by GDPR. The specifics are still being reviewed by regulators, but the underlying question is now firmly on the table.
If you do — or are considering doing — AI labeling work, what exactly are these platforms collecting on you?
What AI Labeling Platforms Know About You
When you sign up for a platform like Outlier or Surge AI, you hand over the obvious: name, email, payment details, and sometimes a copy of your government ID for identity verification. That alone is a significant data point.
But the operational data collected while you work is more revealing.
Keystroke and Session Timing
Most platforms track task completion time down to the second, and some capture fine-grained interaction data: how long you hovered before making a selection, whether you re-read the instructions, how many times you revised an answer. This isn’t necessarily malicious — quality assurance systems use it to flag unusually fast or inconsistent work. But it also builds a detailed behavioral fingerprint.
IP Address and Device Fingerprinting
Your IP address is logged on every task submission. Platforms use this to enforce geographic restrictions, but it also creates a location record of when and where you worked. Device fingerprinting (browser type, screen resolution, installed fonts, timezone) is common as a fraud-prevention measure — but that fingerprint doesn’t expire when the task ends.
Annotation Content
When a task involves labeling photos, rating AI-generated text, or categorizing documents, the content you interact with passes through the platform’s servers. For innocuous tasks like “rate the coherence of these two paragraphs,” this is unremarkable. For tasks involving medical images, legal documents, or personal photos submitted by third parties for quality review, the exposure is more significant.
Payment and Tax Information
In the US, platforms that pay more than $600 in a year require you to submit a W-9 — which includes your full legal name, address, and Social Security Number or EIN. That’s some of the most sensitive personal data that exists, stored in a platform that may have security practices you can’t audit.
The Aggregation Problem
Each of these data points sounds individually manageable. Combined, they’re not.
Your name, verified ID, Social Security Number, home IP address, work schedule, behavioral interaction patterns, and device fingerprint — all in one company’s database — creates a profile that rivals what your bank holds. The difference is that your bank operates under strict regulatory supervision and mandatory breach disclosure laws. Many AI labeling platforms do not.
Smaller platforms in particular may have thin security infrastructure. A breach affecting annotator records wouldn’t just expose an email address. It could expose enough to enable identity theft.
What GDPR and CCPA Actually Cover Here
For workers in the EU or California, the noyb complaint is relevant to watch. If regulators agree that behavioral tracking of platform workers constitutes profiling under GDPR Article 22, platforms would need to offer workers the ability to opt out of automated decision-making that affects their access to work.
The practical concern is task allocation. Platforms use performance scoring — based partly on behavioral data — to determine which workers get access to higher-paying task categories. If that scoring is based on opaque automated profiling, workers may have fewer rights to contest it than they realize.
Under CCPA, California-based workers can request a list of personal information a business has collected about them, ask for deletion, and opt out of the sale of that data. The catch: “sale” under CCPA has a specific definition, and sharing data with a business partner for ad targeting may qualify differently from using it for internal scoring.
Both GDPR and CCPA require businesses to provide a clear privacy notice. That notice should spell out what’s collected, why, and how long it’s kept. Reading it before signing up is worth your time — but the noyb complaint specifically alleges that Scale AI’s consent mechanisms were not straightforward.
Why This Matters Beyond AI Workers
You might not be an AI labeler. The reason this story is worth following anyway is what it reveals about the ecosystem.
The photos, text, and audio files that AI labeling workers process often belong to third parties — users of other apps, customers of other services, or in some cases, scraped from the public internet. Those third parties almost certainly don’t know their content passed through a human labeler on a contract platform in a different country.
This isn’t a hypothetical. AI companies routinely pay contractors to review samples of data flagged by automated systems. Voice recordings, medical images, personal emails in translation tools, legal documents uploaded to AI assistants — all of these can be reviewed by humans as part of quality assurance.
It raises a question that good privacy frameworks are starting to ask: what disclosure do you owe someone when their data passes through a human reviewer, not just an automated system?
Protecting Yourself If You Do This Work
If you currently label data for one of these platforms, or are considering it, some practical steps:
Use a dedicated device or browser profile. Device fingerprinting becomes less useful for cross-context tracking if you separate your labeling work from your personal browsing.
Use a PO box or mail forwarding address. For W-9 and account registration purposes, many platforms accept a valid mailing address that isn’t your home address.
Keep copies of your contracts and payment records. If a platform closes — or is acquired — having your own records of what work you performed and what you were paid matters for taxes and any future disputes.
Read the privacy notice before each new platform. Not the full terms of service, but specifically the privacy notice or privacy policy. Look for how long data is retained, whether it’s shared with third parties, and what your deletion rights are.
Store work-related documents privately. Your W-9 submissions, identity verification confirmations, and payment records are sensitive. Keep those files somewhere you control — not in the same Google Drive folder where you keep everything else.
The Bigger Picture
The labor dynamics of the AI economy are increasingly visible. Major AI labs depend on enormous volumes of human-labeled data. The platforms that organize that labor collect significant personal data about workers in the process — often under terms that don’t fully inform workers of what they’re consenting to.
The noyb complaint against Scale AI is an early sign that regulators are beginning to pay attention. Whether it results in meaningful enforcement or becomes a footnote depends on what investigators find and how platforms respond.
For now, the safest assumption is the same one that applies to any service you sign up for: you are providing data, and that data persists longer and travels further than the signup screen suggests.
Understanding what’s being collected is the first step. What you do with that understanding — which platforms you work with, what information you provide, and how you store your own records — is yours to decide.