/AI1h ago

A vulnerability in Meta's new AI chatbot allowed hackers to compromise over 34,000 Instagram accounts

Hackers hijacked a senior Space Force official's account

20209432034.5K
Gary Marcus#157
Original post
Gary Marcus@GaryMarcus#157inAI

me January 2025 @politico, timing slightly off: we may soon see “the largest cyberattack in history, taking down, at least for a little while, some sizeable piece of the world’s infrastructure… generative AI tools are increasingly being used to create code; in some cases those coders don’t fully understand the code written, and the autogenerated code has already been shown in some cases to introduce new security holes … the world [is] quite vulnerable to ever more audacious attacks.”

meta, right about now: oh shit, he might have a point.

rat king 🐀@MikeIsaac

some news: it turns out 34,000 Instagram accounts got hit by a Meta AI exploit the other day, per internal company docs

hackers also used a senior "Space Force" official's account to post anti-Iran war messaging in the voice of "Hanoi Hannah"

https://www.nytimes.com/2026/06/09/technology/34000-instagram-accounts-ai.html

1:37 PM · Jun 9, 2026 · 3.9K Views
Sentiment

Many users mocked Meta's excuses blaming code errors rather than engineers for the AI exploit that compromised 34,000 Instagram accounts, while expressing sarcasm and fatigue over repeated security failures.

Pos
0.0%
Neg
100.0%
6 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS3.6KBOOKMARKS2RETWEETS3
rat king 🐀@MikeIsaac

20,000 of those accounts didn't have two-factor authentication on, which meant personal info like private messages, phone and email was accessible

3.5k "high value" accounts were stolen before the company says it fixed things for the OG holders

w/ @elitanjourno

1hViews 3.6KLikes 21Bookmarks 2
LIKES24REPLIES4
rat king 🐀@MikeIsaac

@elitanjourno after fixing it, Meta made the point that it was a software bug not endemic to A.I., it just happened to be a bug IN an AI program

which, okay sure.

though i have also heard that line of argument from when an amazon coding bot took out a few of its AWS servers in feb

1hViews 2.7KLikes 24Bookmarks 1
rat king 🐀@MikeIsaac

@elitanjourno you can certainly make the case that "bugs in software are bugs and not AI bugs" fwiw

1hViews 909Likes 3
Jake Halloran@jakehalloran1

@MikeIsaac @elitanjourno yeah just like the aws thing its probably literally true that *AI* is not responsible - their engineers fucking up is

but like still lmao

59mViews 16Likes 1
rat king 🐀@MikeIsaac

@jakehalloran1 @elitanjourno right i think that's the distinction. and like sure, that's fine point to make. i just think sorta have to grit ones teeth to make it

56mViews 15Likes 1
Saroshssnair@saroshssnair

@MikeIsaac Was mythos used?

56mViews 39
John Brownlow@JohnBrownlow

@MikeIsaac @elitanjourno 'It wasn't my mistake, it was a mistake in the code I wrote'

57mViews 6Likes 1
Wash Williams@wash_williams

@MikeIsaac @elitanjourno Wondering if the “software bug” is just the data the AI was trained on being sensitive

1hViews 15
The AI Therapist@TheAIShrink

@MikeIsaac Internal docs confirming the exploit = docs will be subpoenaed. 34k accounts is the headline. Regulatory scrutiny is the actual story.

49mViews 11
AI-Quick-Briefs@aiquickbriefs

@GaryMarcus @politico well mythos is being released to the masses soon so this "doomsday" AI model could spell the end of humanity

45mViews 5
Hunter Gon@gonlenidefi

@GaryMarcus @politico january 2025 me would have eaten this up

now im just tired

53mViews 3
Blissy@BlissyOnX

@GaryMarcus @politico January you was on some oracle mode and now we got Space Force accounts getting hijacked by vintage propaganda

55m
Invincible@InvincibleEdge

@GaryMarcus @politico 2025 prediction finally paying off just late enough that no one will remember u said it

55m