The thing that has always limited cybercrime is people. Skilled hackers are rare, and their time is expensive (which is why most of us have never been individually targeted by one).

AI agents erase that limit. As one expert put it after the OpenAI incident, "Pandora's box is open.”

The tools to run sophisticated, automated attacks at massive scale now exist, and they're getting cheaper by the month.

The same technology that helpfully books your flights can, pointed the wrong way, probe thousands of networks at once without getting tired or making typos.

The silver lining is that defenders get the same superpowers. Much of Black Hat this week is devoted to AI that hunts for security holes before criminals find them, and the OpenAI incident (alarming as it was) was caught, documented, and disclosed, giving the industry a detailed playbook of how these attacks unfold.

For the rest of us, the old advice matters more than ever, because the volume of attacks is about to go up: use unique passwords and a password manager, turn on two-factor authentication, and be skeptical of urgent, too-real-sounding emails and calls.

The "person" behind them might not be a person at all.