Instagram AI‑Powered Support Tools Hijacked, Hundreds of Accounts Locked Out
What Happened — Attackers convinced Meta’s AI‑driven account‑recovery chatbot that they were the legitimate owners of high‑profile Instagram accounts. By feeding AI‑generated facial videos and spoofed geolocation data, they forced the system to change the account‑associated email address, bypassing 2FA and locking out the real owners.
Why It Matters for TPRM —
- AI‑only support flows create a single point of failure that can be weaponized against third‑party assets.
- Credential‑recovery exploits bypass traditional multi‑factor controls, exposing downstream integrations (e.g., marketing APIs, brand assets).
- High‑profile account loss can damage brand reputation and trigger regulatory scrutiny for organizations that rely on these social channels.
Who Is Affected — Social‑media platforms, digital‑marketing agencies, brand‑management firms, public‑figure accounts, and any organization that uses Instagram for customer engagement.
Recommended Actions —
- Review vendor contracts for AI‑assisted support guarantees and escalation clauses.
- Verify that account‑recovery processes include mandatory human review for high‑value accounts.
- Enforce additional authentication factors (hardware tokens, out‑of‑band verification) beyond Meta’s native 2FA.
- Conduct tabletop exercises simulating AI‑driven account takeover scenarios.
Technical Notes — The attack leveraged AI‑generated video avatars to defeat Meta’s selfie verification, exploited a lack of human oversight in the recovery workflow, and used VPN‑based geolocation spoofing to satisfy regional checks. No public CVE is associated; the vulnerability resides in the design of the automated support system. Source: BleepingComputer