Share

Hackers have successfully manipulated Meta’s recently deployed AI customer support chatbot to seize control of high-profile Instagram accounts.

The technology giant confirmed the breach on Monday, stating that it had pushed out an emergency patch over the weekend after security researchers and independent media outlets exposed a catastrophic flaw in the automated verification system.

The security failure compromised a wide array of notable figures, ranging from Barack Obama’s former White House account to the beauty retailer Sephora and U.S. Space Force Chief Master Sergeant John Bentivegna.

Mainstream users also flooded platforms like X and Reddit with complaints of identical hijackings, noting that stolen, highly coveted short handles were quickly put up for resale on Telegram.

The exploit bypassed normal security recovery steps through a shockingly straightforward manipulation of Meta’s automated support assistant, which was rolled out globally earlier this year to handle password resets and mitigate account problems.

According to screenshots and demonstrations circulating among hacking groups on Telegram, attackers initiated the breach by using a Virtual Private Network (VPN) to spoof their geographic location, making it appear as though they were logging in from the victim’s hometown to evade Meta’s localised automated security alerts.

The attacker would then navigate to the Instagram login screen, select the “Forgot Password” prompt, and initiate a chat with the Meta AI support assistant. Instead of providing verified credentials, hackers used targeted prompts to instruct the chatbot to link a brand-new, attacker-controlled email address to the target profile.

The AI bot complied with the request, immediately sending a one-time verification code directly to the hacker’s new email address. Once the attacker pasted the received numbers back into the chat interface, the chatbot displayed a direct link enabling them to reset the password and lock out the legitimate owner.

In a statement addressing the incident, Meta spokesperson Andy Stone confirmed: “This issue has been resolved, and we are securing impacted accounts.” However, the company has not disclosed the total number of users affected by the vulnerability.

The incident raises severe questions regarding Silicon Valley’s aggressive shift toward replacing human customer service with generative AI, particularly for sensitive data administration.

Meta has dramatically scaled its AI infrastructure this year, investing $145 billion in a corporate push toward human-defying “super-intelligence” and automating routine administrative operations.

Cybersecurity experts warn that giving language models direct authority to alter user credentials creates an entirely new frontier for exploitation.

“Companies in every sector are increasingly turning to AI support tools to replace human customer service, and tech giants like Meta are at the forefront of this revolution,” said Marijus Briedis, Chief Technology Officer at NordVPN.

“However, rather than simply providing harmless advice, they are increasingly being given the ability to make real changes to people’s accounts, and that means they need the same level of security checks as a human support agent, if not more.”

Briedis stressed that account recovery is one of the most sensitive elements of any digital platform and should never sacrifice strict verification for convenience. “If an attacker can persuade an automated system to help them bypass normal recovery steps, then the AI becomes part of the attack chain rather than a defence,” Briedis added.