Today, we’re talking with Josh Saxe, an AI security engineer and founder, about what AI actually changes in cybersecurity and what it doesn’t. Josh breaks down who is really behind most attacks, from scam and fraud operations to ransomware groups and nation states, and why even criminals run into the same hiring and adoption hurdles as any other business. He and Adam explore why security has always been a trade-off between protection and friction, and how AI could make strong security far cheaper for defenders. We also get practical about the AI agents already running inside most large companies, including why every agent should have only the access it needs and why letting agents read and act on content from the open web is a decision for the C-suite. So, let’s get into it!

Takeaways:

• Attackers have been slower to adopt AI than many experts predicted, and that gives defenders a window to prepare.
• The groups doing the most damage today are scam and fraud operations, ransomware groups, and nation states. They all hit the same talent and adoption bottlenecks as any other business.
• AI raises the overall level of risk rather than pushing us off a cliff. Security budgets and urgency should go up, but there’s no need to panic.
• Defenders benefit from cheap, widely available AI, because protecting a network takes an enormous amount of AI work, from scanning code to monitoring every device.
• Companies running AI agents need to know which agents they have, give each one only the access it needs, and monitor what they do.
• Security comes down to how much friction an organization is willing to accept. That makes choices like what your agents can read and run a decision for the C-suite and the board.