Instagram announced on Monday, June 2, that it has fixed a critical security vulnerability: a logic flaw in Meta’s own AI-powered account recovery customer service bot allowed attackers to trick the bot into adding a new email address to a target account and triggering a password reset, requiring only the target’s username and no verification through the victim’s original email or password. Last weekend, multiple Reddit and X users reported their accounts being hijacked, including security researcher Jane Huang. The compromised accounts also included high-profile targets: the official White House Instagram account from the Obama administration (deactivated after 2017) and the personal account of U.S. Space Force Senior Enlisted Advisor John Bentivegna.
A leaked attack demonstration video shows the attacker using a VPN to spoof a geographic location matching the target user, then interacting with the Meta AI customer service bot to request linking a new email to the target account. The bot directly sends a verification code to the email provided by the attacker, who then submits the code to gain access to the password reset portal. TechCrunch verified that the email shown in the video indeed received a verification code from Instagram. The vulnerability did not involve any backend server intrusion or database breach; the root cause was that Meta engineers granted the AI customer service bot excessive operational permissions and failed to set rate limits on account operations, allowing black-hat groups to automate bulk operations. Accounts with TOTP two-factor authentication, such as Google Authenticator, were not affected by this vulnerability. Instagram spokesperson Andy Stone confirmed the vulnerability was patched on Monday, but no data on the scale of affected accounts has been disclosed.
BlueDot | IT Home | TechCrunch