What are the hidden dangers lurking beneath the surface of vibe coded apps and hyped-up CEO promises? And what is Influence Ops?I'm joined by Susanna Cox (Disesdi), an AI security architect, researcher, and red teamer who has been working at the intersection of AI and security for over a decade. She provides a masterclass on the current state of AI security, from explaining the "color teams" (red, blue, purple) to breaking down the fundamental vulnerabilities that make GenAI so risky.We dive into the recent wave of AI-driven disasters, from the Tea dating app that exposed its users' sensitive data to the massive Catholic Health breach. We also discuss why the trend of blindly vibe coding is an irresponsible and unethical shortcut that will create endless liabilities in the near term.Susanna also shares her perspective on AI policy, the myth of separating "responsible" from "secure" AI, and the one threat that truly keeps her up at night: the terrifying potential of weaponized globally scaled Influence Ops to manipulate public opinion and democracy itself.
Find Disesdi Susanna Cox:Substack: https://disesdi.substack.com/Socials (LinkedIn, X, etc.): @DisesdiKEY MOMENTS:00:26 - Who is Disesdi Susanna Cox?03:52 - What are Red, Blue, and Purple Teams in Security?07:29 - Probabilistic vs. Deterministic Thinking: Why Data & Security Teams Clash12:32 - How GenAI Security is Different (and Worse) than Classical ML14:39 - Recent AI Disasters: Catholic Health, Agent Smith & the "T" Dating App18:34 - The Unethical Problem with "Vibe Coding"24:32 - "Vibe Companies": The Gaslighting from CEOs About AI30:51 - Why "Responsible AI" and "Secure AI" Are the Same Thing33:13 - Deconstructing the "Woke AI" Panic44:39 - What Keeps an AI Security Expert Up at Night? Influence Ops52:30 - The Vacuous, Haiku-Style Hellscape of LinkedIn
