Your Photos Are Being Reviewed By Strangers on Copilot. There's No Way To Opt Out

Contractors working on Microsoft Copilot are manually reviewing photos uploaded to the AI agent, according to internal documents obtained by 404 Media.^[1] According to a reviewer, the faces aren't blurred.

The documents show that hundreds of contractors see some Copilot users' prompts, uploaded photos, and AI-generated edits, and then rate which edit came out better as part a manual review process.

According to Microsoft’s FAQ, "an opt-out of human review is not available."^[2]

Here’s what the contractors see, why the privacy setting you may have already turned off doesn't cover it, and how to limit what a reviewer can see.

Thanks for subscribing!

Please check your email for your first message.

Take back control of your privacy.

Join thousands who get our privacy-first newsletter every week. No spam. No nonsense.

Sign me up

By submitting this form you agree to receive emails from All About Cookies and to the privacy policy and terms.

Inside the Copilot rating job

The documents 404 Media reviewed describe a rating job and includes instruction guides, real user prompts and images, and posts from an internal contractor message board.  

A contractor sees three things: what a Copilot user typed, the photo they uploaded, and two versions of the edit Copilot produced. The contractor then picks the better edit using four criteria:

• whether it followed the request
• whether the rest of the image stayed intact
• whether it has visible glitches
• how good it looks overall

These workers aren't content moderators. According to 404 Media, their job is to judge quality, and their instructions tell them to trust their gut as a human viewer.

Those same workers see real people's faces. One contractor told 404 Media that "faces are always uncensored" in the images they review. Contractors also described seeing sexual editing requests involving real people, including requests to alter women's clothing and bodies. Many of those requests appeared to be made without the pictured person's consent.

At least one company supplying the workers is Prolific, a firm that recruits people for research and AI feedback work. It didn't respond to 404 Media's request for comment.

A Microsoft spokesperson told the outlet the company uses customer data as described in its terms of use, including to improve its products.

Microsoft isn't alone in employing human reviewers to rate model outputs. Two weeks earlier, 404 Media reported that OpenAI has hundreds of contractors reading real ChatGPT prompts to improve its models. If you've already wondered whether ChatGPT is safe to confide in, the same question now applies to Copilot.

The reporting doesn't show that every Copilot upload gets reviewed. It shows that some users' photos and prompts reached contractors, and there's no way for you to know whether yours did.

What Microsoft's fine print does and doesn't promise

Microsoft's Privacy FAQ for Copilot describes what it does when it prepares uploaded images for AI training. It says it de-identifies them, including by removing metadata and blurring faces. The contractor who spoke to 404 Media described the opposite.

Both accounts can be true. The blurring language sits in the FAQ's section on model training, so it may cover a different process than this rating work. Microsoft hasn't said which applies.

The FAQ is less ambiguous about human review. It says some Copilot conversations get human review "for product improvement and digital safety purposes." It also says human review is required when a Code of Conduct violation is suspected, and that users can't opt out of it. Microsoft hasn't said whether the photo-rating work falls under that policy.

This is where many people could get tripped up. Copilot does let you opt out of having your conversations used for AI model training. That's a separate setting. Nothing Microsoft has published says turning it off keeps human reviewers away from your uploads.

The new app makes the picture murkier. That FAQ applies to the older version of Copilot. Microsoft released an updated app on August 18, 2026. Its privacy pages say prompts, responses, and file contents aren't used to train foundation models, but they don't address human review either way. 404 Media didn't specify which version its documents covered.

You also don't have to use Copilot yourself to end up in front of a reviewer. Your face could be in a photo a friend, coworker, or stranger uploaded. Microsoft's FAQ asks users not to share images of others without their consent, but nothing technically stops them.

Most people don't know any of this happens. In a DuckDuckGo survey of 1,944 U.S. adults, 53% didn't know, or weren't sure, that most chatbots train on users' conversations by default.

People are clear about what they'd want instead. In an All About Cookies survey, 97% of Americans said there should be protections against platforms using their likeness or personal data to train AI. Another 45% said that kind of use should be opt-in only.

How to limit what a Copilot reviewer can see

You can't turn human review off. You can control what a reviewer would find.

1. Upload as if someone might look. Before you share a photo, ask whether you'd be comfortable with a stranger seeing it. Right now, that's the only safeguard that fully works.
2. Keep other people's faces out of it. Ask before uploading photos of friends, family, or coworkers, and be especially careful with photos of kids.
3. Crop before you upload. If the edit doesn't involve anyone's face, like removing a trash can from a backyard shot, crop the face out or pick a different photo.
4. Turn off model training. In the older Copilot app, you can stop your conversations from being used for training. It won't stop human review, but it limits how else your data gets used. Similar settings exist elsewhere, like Meta's AI training opt-out.
5. Delete what you've already shared. Microsoft says uploaded files are stored for up to 18 months, and you can delete conversations anytime. Deleting won't undo a review that's already happened, but it shortens how long your photos sit on Microsoft's servers. Our guide to deleting your ChatGPT history walks through a similar cleanup.

Bottom line

Contractors rating Copilot photo edits see users' faces uncensored, according to one of them, and Microsoft's own FAQ says you can't opt out of human review. Turning off model training doesn't change that. Until Microsoft says otherwise, treat anything you upload to Copilot as something a stranger might see.

Take Control of Your Online Privacy

5.0

AllAboutCookies writers and editors score products based on a number of objective features as well as our expert editorial assessment. Our partners do not influence how we rate products.

Editorial Rating

Claim Deal

On Incogni's website

2026 Editors’ Choice

Best Overall Data Removal Service

Privacy Protection

Incogni

PROMOTION: Get 55% Off Annual Plans With Code COOKIE

• Top-rated data removal service that scrubs your info from 420+ data broker sites automatically
• Independently verified by Deloitte, meaning removals are actually sent, confirmed, and not just claimed
• The Unlimited plan extends coverage to 2,000+ additional sites with human-assisted custom removals

Author Details

Kalleigh Lane

About the Author

Kalleigh Lane is the Managing Editor at All About Cookies, where she helped develop the site's star rating system for product reviews and specializes in testing VPNs, ad blockers, and parental controls on Android and PC. She brings more than five years of experience editing and writing across cybersecurity, tech, and finance, and she has authored and edited over 275 articles for All About Cookies. Before joining AAC, she worked as a journalist and editor at CTV News and the Globe and Mail, and she holds a Master of Arts in English Literature from the University of Toronto.