“It’s never been tougher for a practitioner to secure their environment,” Snehal Antani, CEO at Horizon3, told Cybercrime Magazine at the Black Hat USA 2026 conference in Las Vegas last month.
It’s a new world now where hacking is automated by AI models that don’t have to eat or sleep and that don’t necessarily follow the norms and rules that humans generally do, according to a recent CNET story.
How smart are AI hackers? “These AI hackers are actually kind of dumb,” Antani told us. They look for things they’ve seen before. A human hacker interacts with a decoy 37 percent of the time. An agentic hacker interacts with that same decoy 95 percent of the time.”
AI hackers may be dumb, but they’re extremely dangerous if left unchecked.
“They (AI agents) can escape from the internet and wreak havoc,” said Sydney Von Arx, the chief executive of Nightingale Collective, a not-for-profit organization that helped uncover a recent attack on the RubyGems platform, which was reported by The Wall Street Journal.
Horizon3 has a message for CISOs and security teams: “Go hack yourself, before they do” and “Fight AI with AI,” and it’s resonating because organizations globally now realize that they must safely and autonomously hack their production environment, fix what matters most, verify instantly, and repeat continuously, in order to contend with cyber threats that are now being supercharged by artificial intelligence.
Watch the Video
