Rogue AI agent prompts Meta employee to leak sensitive data

A "large amount" of sensitive user data was internally exposed by an artificial intelligence agent at Facebook's parent company, Meta, according to the Guardian.

What's happening?

A Meta engineer using agentic AI — artificial intelligence capable of performing multi-step tasks — was instructed to perform an action that inadvertently exposed private company and user data.

The unnamed individual sought "guidance on an engineering problem" via an internal system, according to the Guardian, in which the AI agent provided the solution, leaving the data exposed for two hours.  

A spokesperson for Meta claimed "no user data was mishandled" and maintained that human employees could also inadvertently provide bad instructions. 

But, as the Guardian noted, the incident was not an isolated one: Other major players in the AI sector have experienced similar internal breaches and breakdowns caused by AI agents.

The outlet cited its recent reporting on a similar incident at Amazon. Several insiders at that company said its efforts to bring AI operations into virtually all Amazon work had been disastrous.

Security specialist Jamieson O'Reilly told the Guardian that AI agents lacked the ability to recall "context" as human programmers do.

"A human engineer who has worked somewhere for two years walks around with an accumulated sense of what matters, what breaks at 2 a.m. … That context lives in them, in their long-term memory, even if it's not front of mind," O'Reilly said. 

Why is this concerning?

AI consulting firm co-founder Tarek Nseir made a similar observation to the outlet about the "experimental" nature of these incidents.

Firms like Meta are "not really kind of standing back from these things and actually really taking an appropriate risk assessment," he began.

Which of these savings plans for rooftop solar panels would be most appealing for you? Save $1,000 this year 💸 Save less this year but $20k in 10 years 💰 Save less in 10 years but $80k in 20 years 🤑 Couldn't pay me to go solar 😒 Click your choice to see results and earn rewards to spend on home upgrades.

"The vulnerability would have been very, very obvious to Meta in retrospect, if not in the moment. And what I can say and will say is this is Meta experimenting at scale. It's Meta being bold," Nseir added.

Meta's desperate push to come out ahead on AI is well-documented, and the leak coincided with the high-profile failure of the Metaverse, a virtual reality project that cost as much as $80 billion. 

As Silicon Valley's biggest names continue pouring money into a technology with decidedly underwhelming capabilities to date, AI and its data centers have become increasingly controversial.

Data centers consume power and water at high rates, and while the facilities themselves have been described as a local nuisance, the electric bills resulting from their immense energy demand are spread out far more widely.

Despite growing public backlash, however, Meta was given the green light to double its AI investment in 2026.

What's being done about it?

According to TechCrunch, Meta considered the breach severe, and it came on the heels of another agentic AI glitch at the company.

As for AI and data centers, communities have increasingly stood up to data firms, blocking $98 billion in new data center development in the final quarter of 2025.

Get TCD's free newsletters for easy tips to save more, waste less, and make smarter choices — and earn up to $5,000 toward clean upgrades in TCD's exclusive Rewards Club.