Meta's AI Just Stole $1M

Summary

The video discusses a recent exploit where hackers used social engineering on Meta's AI assistant to steal high-value Instagram accounts. The hosts explain the attack vector, including prompt injection and bypassing 2FA, and criticize Meta's security response. They also offer advice on personal account protection and highlight broader AI security risks.

• Hackers exploited Meta's AI account recovery assistant to reset passwords for valuable Instagram handles.
• The exploit involved social engineering and prompt injection rather than code vulnerabilities.
• Two-factor authentication was bypassed through developer mode impersonation on Facebook.
• Meta's response claiming no breach was criticized as insufficient.
• The hosts discuss the 'confused deputy' problem in AI security.
• Personal protection tips include hardware security keys and avoiding SMS 2FA.
• The exploit highlights a growing attack vector as AI systems become more integrated.
• The White House released a mandate for proactive AI cybersecurity measures.