The Meta hack shows there's more to AI security than Mythos

Attackers successfully used Meta's AI customer support agent to steal Instagram accounts, including the dormant Obama White House account, by simply asking the AI to transfer accounts to attacker-controlled email addresses. The incident highlights how AI systems can become attack vectors through social engineering, even when they're not directly compromised. This breach demonstrates that AI security requires more than just protecting the models themselves—it demands careful consideration of how AI agents interact with sensitive systems and data.