Hackers Exploit Meta’s AI Chatbot to Hijack Instagram Accounts
In a recent security breach, hackers successfully commandeered several Instagram accounts by manipulating Meta’s AI-powered support chatbot. This sophisticated attack exploited the chatbot’s functionalities to gain unauthorized access to user accounts.
Over the weekend, numerous Reddit users reported unauthorized access to their Instagram accounts. Similarly, several users on X (formerly Twitter) raised alarms about account hijackings. Notably, compromised accounts included the inactive Instagram handle of the Obama-era White House and the account of U.S. Space Force’s Chief Master Sergeant John Bentivegna.
Security researcher Jane Wong also fell victim to this attack. She reported that her Instagram password was changed without her knowledge and that she received multiple password reset attempts. Wong expressed her concerns, stating, The password got changed without my knowledge and I was getting different password reset attempts throughout yesterday. Quite concerning.
A video circulating on X detailed the hackers’ method. The attackers used a VPN to spoof the target’s location, thereby bypassing Instagram’s automated security measures. They then initiated a chat with Meta’s AI Support Assistant, requesting the addition of a new email address to the victim’s account. The chatbot sent a verification code to the hacker’s provided email, which the hacker then relayed back to the chatbot. This process prompted the chatbot to display a Reset Password option, allowing the hacker to set a new password and gain control over the victim’s account.
TechCrunch verified the authenticity of this method by confirming that the hacker’s public email received the verification code as demonstrated in the video. Crucially, at no point did the hacker need access to the legitimate email associated with the victim’s Instagram account.
On Monday, Instagram spokesperson Andy Stone addressed the issue, stating that the vulnerability had been fixed. However, the exact number of affected users remains unclear. Meta has yet to provide an official comment on the incident.
