The hackers didn't have to crack a single password—they simply asked Meta’s AI support to change the email addresses associated with other people's Instagram accounts.
The hackers didn’t have to crack a single password—they simply asked Meta’s AI support to change the email addresses associated with other people's Instagram accounts, and the bot did it. Over 20,000 accounts were affected, including the White House's account; there was simply no provision for human intervention. In WeeklyTalk, we use this and other cases to show how the old help desk trick becomes a new vulnerability as soon as an AI is allowed to act independently.
And we ask what this means for critical processes—where everything hinges on one question: Who is actually authorized to make changes?
