Hackers Exploit Meta’s AI Chatbot to Hijack High‑Profile Instagram Accounts Including Obama White House and Sephora

Meta’s push to automate customer support with artificial intelligence has come under scrutiny after hackers claimed they were able to hijack Instagram accounts by manipulating the company’s AI support chatbot. The alleged flaw, which has now been fixed according to Meta, reportedly enabled attackers to reset account credentials and gain control of profiles without ever accessing the victim’s email account.

What Happened?

According to reports shared by security researchers and affected users, attackers discovered a way to trick Meta’s AI Support Assistant into changing the email address linked to an Instagram account.

The process was alarmingly simple. Hackers allegedly initiated a chat with Meta’s support bot and requested that a new email address be linked to a target Instagram profile. The chatbot then sent a verification code directly to the email address supplied by the attacker. Once the code was entered, the AI assistant reportedly offered a password reset option, allowing the attacker to create a new password and take control of the account.

Critically, the attack did not require access to the victim’s original email address, a safeguard that is typically central to account recovery systems.

High-Profile Accounts Reportedly Targeted

The alleged exploit coincided with a wave of Instagram account takeovers involving several well-known profiles.

Among the accounts reportedly compromised were the Instagram handle associated with the Obama-era White House, beauty retailer Sephora, and the account of US Space Force Chief Master Sergeant John Bentivegna.

Security researcher Jane Wong also revealed that her Instagram account had been taken over, saying her password was changed without her knowledge while multiple password reset attempts were being made.

How the Attack Worked

Videos circulating on social media and private security channels appeared to demonstrate the attack step by step.

Researchers said attackers first used a VPN to mimic the target’s geographic location, potentially reducing the chances of triggering Instagram’s automated security systems. They then contacted Meta’s AI support chatbot, requested an email change for the target account, received the verification code at their own email address, and used it to initiate a password reset.

The apparent weakness lay in the chatbot’s failure to adequately verify ownership of the account before allowing critical security changes.

Meta Says Issue Has Been Fixed

Following growing reports of account hijackings, Instagram spokesperson Andy Stone stated that the company had identified and resolved the security issue.

Meta has not disclosed how many users may have been affected before the flaw was patched. The company also has not provided detailed technical information about how the vulnerability bypassed existing account protection measures.

Bigger Questions About AI-Powered Support

The incident highlights the risks of giving AI systems authority over sensitive account recovery functions.

Earlier this year, Meta expanded AI-driven customer support across Facebook and Instagram, promoting features that could handle password resets, account recovery, and other security-related requests. While automation can reduce response times and operating costs, critics argue that critical account functions still require strong human oversight.

For affected users, another concern remains the difficulty of reaching a human support representative when an account is compromised. Many users have complained that AI-powered support channels often provide limited escalation options during security emergencies.

A Warning for Tech Companies

The episode serves as a reminder that AI assistants can become attractive targets for attackers if they are granted access to sensitive account controls. Even sophisticated security systems can be undermined when automated support tools fail to verify identity properly.

As technology companies increasingly deploy AI to handle customer service and account management, experts say security safeguards must evolve just as quickly to prevent convenience from becoming a new avenue for cyberattacks.

Read More: Meta Wants You to Pay for Instagram, Facebook and WhatsApp Now With 'Plus' Subscriptions