Loading...
Loading...
A Tell HN post alleges a critical flaw in Meta’s AI-powered account support for Instagram that lets attackers hijack accounts. The poster says the feature—A/B tested for a subset of users—allows an attacker using a proxy or VPN near the target’s region to request a verification code be sent to any email, relay the code to the AI agent, and obtain a password-reset link to take over the account. The reporter claims over 100 high-value accounts were compromised and urges Meta to disable the AI supp
AI-driven account recovery is being used in the wild to bypass traditional defenses, creating a new attack vector that can impact high-value accounts and trust in platform security. Tech professionals must reassess AI-assisted support flows and incident response controls to prevent large-scale account takeovers.
Dossier last updated: 2026-06-01 19:28:22
Hackers claim Meta’s AI customer-support chatbot was used to take over high-profile Instagram accounts by instructing the bot to change the accounts’ recovery email addresses. Videos and screenshots shared in Telegram groups reportedly show attackers prompting the AI to link a new email and then submitting verification codes, enabling control of targets including notable public and corporate profiles. The incidents highlight risks from delegating critical account-recovery functions to automated agents after Meta rolled out AI-driven support across Facebook and Instagram, which can reset passwords and perform account maintenance. Affected users report limited options to reach human support, raising concerns about security, escalation paths, and abuse of AI-enabled workflows.
Instagram suffered a simple but wide-reaching account takeover flaw where attackers used only a target username and a VPN to spoof location and convince Meta's AI support to send verification codes to attacker-controlled emails. Reportedly the AI would accept arbitrary email addresses and in some cases weak video-selfie proofs (including AI-animated public photos), allowing attackers to reset passwords, revoke 2FA, and seize accounts without triggering owner notifications. High-profile accounts including the Obama White House and a U.S. Space Force official were affected; criminal marketplaces on Telegram offered takeover services. Meta has reportedly patched the issue, but the incident highlights weak guardrails in automated support flows and the risks of AI-driven account recovery.
A wave of Instagram account takeovers, including high-profile profiles such as the Obama White House account, was reportedly enabled by a flaw in Meta’s AI-driven support recovery flow. The article describes an attack requiring only a username: attackers spoof a victim’s region using a VPN/proxy, then tell Instagram’s support AI the account is hacked and request verification codes be sent to an email address they control. With the code, they obtain a password reset link and take ownership. The process can bypass two-factor authentication, revoke existing sessions, and change recovery email/phone without notifying the original owner. Telegram groups allegedly sold takeover services, targeting valuable short handles. Meta appears to have patched the issue after it may have persisted for weeks or months.
A Tell HN post alleges a critical flaw in Meta’s AI-powered account support for Instagram that lets attackers hijack accounts. The poster says the feature—A/B tested for a subset of users—allows an attacker using a proxy or VPN near the target’s region to request a verification code be sent to any email, relay the code to the AI agent, and obtain a password-reset link to take over the account. The reporter claims over 100 high-value accounts were compromised and urges Meta to disable the AI support feature and revert hijacked usernames. The post also notes a prior unacknowledged Instagram data-exposure issue and expresses concern the bug may be actively exploited in cybercriminal channels.