In this episode of Unspoken Security, host AJ Nash sits down with Snehal Antani, co-founder and CEO of Horizon3.ai and former first CTO of Joint Special Operations Command (JSOC), to dissect what AI is really doing to the economics of cyber attacks - and cyber defense. Snehal opens with a jaw-dropping data point: his team compromised a defense industrial base supplier and achieved full domain admin in 77 seconds. From there, the conversation moves into the surprising counter-strategy his team uncovered - that today's LLMs and agentic attackers are dramatically more gullible than human hackers, clicking on well-placed honey tokens up to 95% of the time.

The two dig into why "train like you fight" - a principle Snehal absorbed at JSOC - is now essential for cyber teams, why compliance frameworks like CMMC are failing to produce real resilience, and why the future of cyber warfare is "AI versus AI with humans by exception." Snehal walks through Horizon3.ai's architectural bet on disposable models, persistent knowledge graphs, and constrained-action-space agents, and explains why the "haves and have-nots" of red teaming is finally being disrupted by autonomous pentesting that IT admins - not elite ethical hackers - can operate.

The conversation closes with a candid look at the industry itself: the ChatGPT-driven sameness of vendor messaging, the Black Hat gimmick arms race, and Snehal's plea to return to technical authenticity. He ends with a deeply personal reflection on his late father - the electrical engineer who sabotaged toy robots so his six-year-old son would learn to troubleshoot them - and the weight of trying to pass that same gift on to his own kids.