Meta’s AI support chatbot is handing Instagram accounts to hackers
In June 2026, hackers discovered something alarming: Meta’s AI support chatbot does not verify your identity. They asked to change the email addresses on the Instagram accounts. The chatbot complied. Within minutes, high-profile accounts, the Obama White House profile, the US Space Force Chief’s account, Sephora’s verified handle, and security researcher Jane Wong’s profile, were compromised.
No passwords cracked. No phishing. Just a conversation with an AI that had no business having the power it was given.
How It Works
The attack is straightforward. Hackers use a VPN to spoof a target’s location, bypassing geolocation warnings. They access Meta’s “Get Support” chatbot and select “Forgot Password.” The AI asks basic questions. Hackers claim to be the account owner. The AI, programmed to be helpful, asks for verification but accepts any answer. When hackers request an email change, the chatbot obliges. An automated one-time passcode goes to the hacker’s email. They enter the code. Instagram’s security system treats this as a legitimate account recovery. Password reset. Lockout complete. The legitimate owner sees nothing until it is too late.
Why Meta Made This Mistake
In March 2026, Meta announced 24/7 AI support for Facebook and Instagram. The company positioned it as innovation: “Solutions, not just suggestions.” What Meta actually did was wire elevated permissions directly into an AI chatbot. The system could reset passwords, change email addresses, and modify critical account settings, all without human verification.
This was not a security oversight. It was a business decision. Human support costs money. AI support scales infinitely. Meta chose scale over security.
The Pattern of Failure
Meta said the issue was fixed in early June. By Tuesday, hackers were still exploiting it. Screenshots and videos circulated in Telegram hacking communities showing successful takeovers happening after the alleged patch. Meta scrambled to alert victims. But there was no escalation path. Victims could not contact humans. The same AI that locked them out was now their only support option.
What This Costs
This is not just a security failure. It is a brand collapse. Instagram users trusted Meta with their most valuable digital asset: their identity, their audience, and their verified status. Meta repaid that trust by making accounts easier to steal than to protect.
Brand equity is not built on features. It is built on the belief that a company will not betray you. Meta betrayed that belief by prioritising AI convenience over human accountability.
When victims cannot escalate to humans, they are not talking to a company. They are talking to a system that does not care. That is not a brand experience. That is abandonment.
The Deeper Problem
The real issue is not that Meta’s AI is dumb. It is that Meta gave it access that it should not have. AI chatbots are excellent at answering questions. They are catastrophic at making decisions about account security. Meta confused capability with responsibility.
Every account takeover going forward will be a reminder: Meta chose efficiency over trust. That choice has a cost, one measured not in stolen accounts, but in eroded brand loyalty.
