Can AI Agents Actually Hack Systems?

A new AI is being framed as a tool that can find zero-days fast and even “hack its way out” of containment. If that claim is real, defenders, developers, and everyday users are about to feel the consequences. On this episode of Legitimate Cybersecurity, hosts Frank Downs and Dustin Brewer are joined by Jason Casey, CEO of Beyond Identity, to break down the panic around Anthropic’s “Mythos” discussion, what AI can actually do for offense and defense, and where the marketing may be outrunning the real-world risk. They dig into whether this is a true cybersecurity turning point, or the latest example of the industry turning fear into momentum. They also explore how AI is already reshaping blue team work, governance, detection, and security operations. Plus: hacked smart vacuums, trackable e-ink nails, wearable surveillance, and why convenience keeps creating new attack surfaces nobody asked for. Media/interview: [email protected] Audio: https://legitimatecybersecurity.podbean.com/ Chapters: 00:00 The new AI panic begins 00:59 What “Mythos” is supposed to do 02:17 Is this a real threat or brilliant marketing? 07:12 Will this change security budgets and priorities? 10:11 Why cybersecurity leaders amplify moments like this 13:58 How AI actually helps blue teams 21:49 Rules, patterns, and better AI detection 23:59 The idea of an AI “security factory” 31:50 Beyond Identity’s new governance layer 35:30 Hacked vacuums, smart nails, and wearable tracking 45:00 Final takeaways #legitimatecybersecurity #artificialintelligence #cybersecurity #anthropic #claude #aisecurity #zerodayjay #blueteam