Meta AI Duped! Hackers Weaponize Support Chatbot to Hijack High-Profile Instagram Accounts

 

A catastrophic security flaw has exposed the dark side of artificial intelligence automation. In a shocking cybersecurity breach confirmed by industry insiders, hackers successfully manipulated Meta’s AI-powered customer support assistant to hijack several high-profile Instagram accounts. The compromised pages reportedly include the dormant Obama White House account, global beauty giant Sephora, and even the personal account of a senior U.S. Space Force official.

The high-profile incident has wiped over five percent off Meta’s stock value and triggered intense scrutiny over Big Tech's rush to replace human security teams with automated AI bots.

How Hackers Outsmarted Meta’s AI: The "Prompt Injection" Exploit

According to security researchers and leaked screen recordings shared on Telegram and X, the breach was executed using a sophisticated technique known as a "prompt injection" attack. Instead of using complex coding to bypass Meta’s two-factor authentication, attackers simply held a text conversation with the Meta AI support chatbot. 

By tricking the AI into believing they were the rightful owners attempting emergency account recovery, the hackers persuaded the bot to link the targeted Instagram profiles to completely new, attacker-controlled email addresses. The AI chatbot then sent a verification code to the new email and openly displayed a password reset link directly inside the chat window—completely bypassing standard password safeguards.

The Aggressive AI Push Backfires on Mark Zuckerberg

This security disaster comes at a highly sensitive time for Meta. Under CEO Mark Zuckerberg, the social media titan has been aggressively restructuring, cutting thousands of human support roles while pledging up to $145 billion toward artificial intelligence infrastructure. 

The strategy was designed to let AI handle repetitive tasks like account recovery and content moderation. However, cybersecurity experts tell Reuters that this high-profile breach proves AI large language models (LLMs) are still far too unstable and easily manipulated to be trusted with high-level security permissions.

Stolen Digital Handles Sold on Black Market

While security researchers like Jane Wong (a former Meta employee whose own handle was caught in the crossfire) managed to recover access within minutes, thousands of ordinary users weren't as lucky. Reports indicate that multiple hijacked premium Instagram handles were immediately put up for auction on underground Telegram hacking channels before Meta could patch the vulnerability.

Meta's Official Response and Future Outlook

In an official statement addressing the crisis, Meta stated, "This issue has been resolved, and we are actively securing all impacted accounts." However, the company declined to release the exact number of accounts compromised during the multi-day exploit. 

While the patch stops this specific chatbot loophole, tech analysts warn that the underlying vulnerability of automated AI systems remains a ticking time bomb for the entire tech industry.