Second Instagram flaw exposed Zuckerberg’s contact info days after AI chatbot hack

Meta AI Chatbot Exploited to Hijack High-Profile Instagram Accounts

Hackers exploited Meta’s AI-powered support chatbot to seize control of prominent Instagram accounts using a prompt injection attack, while a separate logic bug in the platform’s web password reset flow briefly exposed unredacted personal data tied to accounts including that of Meta CEO Mark Zuckerberg.

The AI Chatbot Takeover

The attack, which came to wide public attention in late May and early June 2026, required remarkably little technical sophistication. As first reported by 404 Media, attackers initiated conversations with Meta’s AI support assistant and simply instructed it to link a new email address to a targeted account. The bot complied, sending a verification code to the attacker’s inbox and then facilitating a full password reset — bypassing standard identity verification and, in many cases, two-factor authentication.404media

Among the accounts compromised were the Obama-era White House archive page, beauty retailer Sephora, U.S. Space Force Chief Master Sergeant John Bentivegna’s account, and cybersecurity researcher Jane Manchun Wong’s handle. Stolen accounts were reportedly resold on Telegram channels, with the combined market value of hijacked handles estimated at over $1 million.aiweekly

Meta spokesperson Andy Stone confirmed on X that the vulnerability had been resolved: “This issue has been resolved and we are securing impacted accounts”. The company deployed an emergency patch that disabled or restricted the AI chatbot’s direct write access to email-binding and password-reset APIs.cybersecuritynews

A Second Flaw Exposes Zuckerberg’s Contact Details

On June 6, a related but distinct logic bug was discovered in Instagram’s web-based password reset interface. Security researcher @Scot0xo publicly demonstrated that initiating a standard password reset for any username returned fully visible email addresses and phone numbers rather than the partially redacted versions Instagram normally displays. Proof-of-concept screenshots shared by @vxunderground showed the login screen for the account “zuck” revealing multiple associated emails and a linked phone number.x

Meta deployed an emergency hotfix within hours of the disclosure but not before the demonstrations circulated widely on social media.cybersecuritynews

Systemic Concerns

Security researchers have pointed to a common thread connecting the incidents: architectural decisions that grant AI systems privileged access to account management functions without deterministic authorization checkpoints. According to KrebsOnSecurity, the prompt injection attack was unsuccessful against accounts with multi-factor authentication enabled. Reuters reported that the breach has raised alarms about Meta’s broader strategy of automating sensitive user operations.reuters

Open-source intelligence researcher ZachXBT described the flaw bluntly: “Meta AI’s support is rubbish; it has tons of access privileges, yet it can reset the password for any user’s account without two-factor authentication, and it doesn’t even verify who you are”. Meta has not disclosed a CVE identifier for either vulnerability.gigazine