Instagram security flaw: How hackers tricked Meta’s AI into handing over accounts – Firstpost | #hacking | #cybersecurity | #infosec | #comptia | #pentest | #hacker

Meta has recently patched a security vulnerability that allegedly allowed attackers to take control of Instagram accounts by exploiting the company’s AI-powered support chatbot, following a wave of high-profile account hijackings that unfolded over the weekend.

The issue came to light after several prominent Instagram profiles were compromised, including the account associated with the Obama-era White House, which has more than two million followers. The account, largely inactive since 2017, briefly displayed a politically charged message claiming that the White House was under the control of Shiites. Another affected profile belonged to Chief Master Sergeant of the US Space Force, John Bentivegna.

Here is the full method:

(How to hack Instagram accounts)

Disclaimer: this video was posted on Telegram, and I have not attempted to use it.
The purpose of this post is to show you how @AIatMeta sucks and how AI is falsely disabling accounts & has serious problems like these. https://t.co/1N98XmwvRV pic.twitter.com/rVRw7t39Ry

— Bahram Sahbani (@bahrambiz) May 31, 2026

How did hackers attack Instagram accounts?

As screenshots and videos documenting the attacks spread across social media platforms, cybersecurity observers and online researchers began piecing together how the intrusions were carried out. Rather than relying on sophisticated malware or traditional hacking techniques, the attackers appeared to exploit a weakness in Meta’s own AI-driven customer support system.

According to videos and screenshots circulating online, attackers initiated conversations with Meta’s AI Support Assistant and convinced the chatbot that they were authorised users seeking access to specific Instagram accounts. The chatbot was allegedly persuaded to add a new email address, controlled by the attacker, to the targeted account.

meta gave their AI support agent the ability to modify your instagram account. no identity verification. people figured this out and accounts are being taken over right now pic.twitter.com/60yRrImnaZ

— impulsive (@weezerOSINT) May 31, 2026

Once the new email address was linked, the chatbot reportedly sent a verification code to that address. Attackers then supplied the code back to the AI assistant, which provided access to password reset options. By setting a new password, the attackers were able to take over the account entirely.

The method effectively turned a common social engineering tactic into an AI-powered exploit. Instead of manipulating a human support representative, attackers appeared to manipulate an automated system that failed to properly verify account ownership before carrying out sensitive requests.

Several of the demonstrations showing the exploit were reportedly shared through Telegram channels where cybercriminals trade hacking methods and digital tools. Other users claimed they independently reproduced the technique, adding to concerns that the vulnerability could be widely abused.

What made the flaw particularly alarming was that account owners had little ability to defend themselves. Reports suggest the chatbot’s actions may have bypassed standard security measures, including safeguards that would normally prevent unauthorised password changes. As a result, even users with additional account protections in place could have been vulnerable.

The incident highlights the growing challenges technology companies face as they integrate artificial intelligence into customer service and account management systems. While AI assistants can help automate routine support tasks and reduce response times, security experts have repeatedly warned that these systems can become targets for manipulation if safeguards are not rigorously tested.

Meta has not publicly detailed the technical cause of the vulnerability, but the company appears to have acted quickly after reports gained traction online. Since the exploit became widely known, users attempting to replicate the method have reported that it no longer works, indicating that the loophole has been closed.

The episode serves as a reminder that as AI becomes more deeply embedded in online platforms, companies must ensure convenience does not come at the expense of security.