Multi-agent cyberattacks are lowering the cost of exploitation

A reported campaign used AI agents against roughly 100 online retailers, allegedly stealing about 600,000 cards at roughly $25 per target, while related reporting describes rapid RCE, domain-admin access, and multi-organization compromise. Verification, containment, attack-path chaining, and the durability of model-safety refusals remain open questions as agents automate more of the kill chain.