Instagram’s AI Support Bot Exploited to Hijack User Accounts
Meta has patched a serious vulnerability in its Instagram platform after BBC Business reported Tuesday that hackers manipulated an AI support chatbot to seize control of other users’ accounts.
How the Exploit Worked
Attackers reportedly used Instagram’s built-in Meta AI assistant to reset passwords on accounts they did not own. The method involved masking their real location with a virtual private network to mimic the target account holder’s location. They then asked the AI tool to attach a new email address to the target account. The chatbot complied, forwarding a verification code to the attacker’s inbox. That code unlocked a password-reset link, effectively handing over the account.
Meta spokesperson Andy Stone confirmed the fix on X, stating that the issue had been resolved and that compromised accounts were being secured. Stone separately dismissed reports that world leaders’ accounts had been targeted via the same method, calling those claims false.
High-Profile Accounts Among Those Reportedly Affected
Tech outlet 404 Media noted the timing of disclosed exploit activity coincided with a cluster of notable account takeovers. One affected profile was an Instagram account associated with former US President Barack Obama’s White House tenure. That account reportedly published pro-Iran content before it was recovered. The scope of affected accounts beyond high-profile cases remains unclear.
Security researcher and former Meta engineer Jane Manchun Wong said her own Instagram password was changed without her knowledge. She described receiving repeated unauthorised password-reset attempts and called the experience “quite concerning.”
A Recurring Problem With AI-Driven Support
The incident lands against a backdrop of growing unease over AI systems handling sensitive user data. Critics point out that removing human oversight from account recovery processes creates dangerous blind spots. One affected user wrote publicly that an AI had stolen their account while a separate AI could not recover it, noting there were no humans involved at any stage.
Meta has faced mounting criticism over its user support infrastructure. An independent EU dispute body said last week that Meta almost never responds when it raises cases involving wrongful account bans on Facebook and Instagram. The company has also made significant workforce reductions while simultaneously pouring billions into AI development.
Also Read: WhatsApp Defends ‘Optional’ AI Tool That Cannot Be Turned Off
What Comes Next for Meta’s AI Rollout
Pressure is building on Meta to reintroduce human support pathways for account recovery. The exploit highlights a structural risk in deploying generative AI tools for authentication-adjacent tasks without sufficient safeguards. Regulators and security researchers are likely to scrutinise how the flaw went undetected and for how long it was active.
Read Next: Meta Repeatedly Snubs EU Body Over Facebook and Instagram User Bans
