A Meta engineer did not want their own screen recorded for AI. Now Meta wants to read your inbox.

There is a story from inside Meta this year that keeps coming back to me, and this week it feels more relevant than ever.

What happened inside Meta

In April 2026, Meta rolled out something called the Model Capability Initiative. It recorded employees' mouse movements, clicks, keystrokes, and navigation on their work computers, so Meta could use that data to train AI agents to operate a computer on their own. Meta's CTO, Andrew Bosworth, confirmed employees could not opt out, saying the company needed "real examples of how people actually use computers." (Yahoo Tech)

A veteran Meta engineer, who has not been publicly named, wrote an internal post that nearly 20,000 employees viewed. The key line:

"Selfishly, I don't want my screen scraped because it feels like an invasion of my privacy. But zooming out, I don't want to live in a world where humans, employees or otherwise, are exploited for their training data." (Gadget Review)

Notice "or otherwise." That is not just about employees. That is about everyone.

Then it got worse

The backlash spread. Flyers showed up in Meta's California and New York offices. Over 1,600 employees signed a petition, and one office nicknamed the program the "Employee Data Extraction Factory." (The Next Web)

Meta added more opt-out options a few weeks after launch. Then in June it paused the program entirely after the keystroke and mouse data ended up readable across the company. Bosworth said a researcher "had put it in a place it wasn't supposed to go," and stressed there was no outside breach. What Meta plans to do with the data it already collected is still unanswered. (The Next Web)

So to recap: Meta collected highly sensitive data from its own people, made it mandatory, and then lost track of where it went internally. That is the company's own staff, with every advantage of being inside the building.

Now look at Muse

On September 8, Meta launched Muse, a personal AI agent. To be useful, it asks for access to your email, your calendar, your payments, and your shopping accounts. It has been a hit, number one on the App Store. (GeekWire)

Meta says Muse runs each session in an isolated virtual machine and never sees your actual passwords. (Tech Times) Maybe that design holds up. But within two weeks, a researcher showed Muse could be tricked by a malicious web page, and Amazon blocked it from shopping on its site, alleging it stores customer credentials. (TechCrunch) Meta disputes that.

To be precise about what the engineer said

I want to be careful here, because the internet has turned this story into something slightly different. The engineer did not publicly tell people to stop using Facebook, Instagram, or WhatsApp. The post was about being recorded for AI training at work.

But you do not need an engineer to tell you to delete your apps to learn something from this. When the people who build a company's products push back this hard on how that company treats their data, that is information. And the point they made, that nobody should be "exploited for their training data," applies just as much to you as to them.

What I would actually do

I am not going to tell you to delete every Meta app. Plenty of people need WhatsApp for family abroad or Facebook groups for their community. Here is a more practical approach:

1. Keep Meta's AI agent away from your accounts. Using Instagram is one thing. Giving an AI your email and shopping logins is a much bigger level of trust. I would wait until Muse has a longer track record.
2. Check your AI training settings. Meta's AI features can use your interactions to improve its models. Look in each app's privacy settings and turn off what you can.
3. Use Meta apps for what they are good at, and nothing more. Messaging and photos, fine. Your health, money, and private notes belong somewhere else.
4. Prefer tools that do the work on your device. The less of your life that goes to a server, the less there is to misplace.

That last one is why I build my apps the way I do, and it is why I pay attention to stories like this one. The tech is not the problem. The question is always who holds your data and what they are allowed to do with it.

Verified September 22, 2026. The internal post and program details come from published reports citing Meta employees and executives; the engineer has not been publicly identified. Amazon's and Meta's claims about Muse are unproven. The studio's app lineup is at jcmobileappstudio.com/apps.

Sources: Yahoo Tech, Gadget Review, The Next Web, GeekWire, TechCrunch, Tech Times