#ai-security
17 episodes
#5481: Small Models, Big Guardrails: PII Detection in ChatGPT
A tiny 600M parameter model is quietly scanning your ChatGPT tool calls for PII. Here's how that class of guardian model actually works.
#5434: When Local AI Tools Phone Home
Local AI tools shouldn't need the network. So why is your on-device writing assistant uploading 40MB at 3am?
#5404: Gemini Broke Out of Its Sandbox. Sort Of.
A Gemini agent reached three real companies during a capture-the-flag test. The containment failure, the seven-week silence, and what "broke out" a...
#5113: Can Rival Labs Poison AI Training Data?
A few hundred crafted documents can shift a model's stance. Here's how data poisoning actually works.
#3422: How Rival Labs Reverse-Engineer a New AI Model in Hours
Inside the organized frenzy when a closed-source model drops — and how competitors map its every weakness.
#2691: The Usability Tax of Least Privilege
Is it time to let AI agents handle your API key creation and rotation? We explore the real security tradeoffs.
#2482: When AI Chatbots Leak Your PDFs via Public S3 Buckets
A user uploaded a sensitive PDF to an AI chatbot. The chatbot stored it in a public S3 bucket with zero authentication.
#2472: When Guardrails Break: The Hidden Costs of AI Gateway Filtering
PII detection at the gateway layer can block legitimate invoices. Here's how guardrails actually work and where they fail.
#2180: The Sandboxing Tradeoff in Agent Design
AI agents need broad permissions to be useful—but every permission expands the attack surface. We map the real threat landscape and the isolation t...
#1905: How VCs Verify AI Startups Without Stealing Code
From the "No-NDA Paradox" to AWS bill forensics, here’s how investors separate real AI from Raspberry Pis in fancy cases.
#1474: The End of API Keys: Securing Non-Human Identity
Stop leaving your digital keys under the mat. Learn how workload identity federation is replacing the dangerous "secret management grind."
#1235: Beyond "No Training": Securing the New Agentic AI Stack
Think your data is safe because of a "no training" clause? We deconstruct the hidden security risks within the modern agentic AI stack.
#1217: The Missing Ring Zero: Why LLMs Can't Keep Secrets
Discover why AI models leak their secret instructions and how to defend your intellectual property using modern prompt hardening techniques.
#679: The Sound of Secrets: Side-Channel Attacks in AI Clusters
Is your hardware whispering your secrets? Discover how side-channel attacks turn physical signals into data leaks in modern AI clusters.
#671: Keys to the Kingdom: Securing AI Model Weights
How do AI labs share their models without losing the secret sauce? Explore the tech keeping Claude secure in the Pentagon’s hands.
#168: The Sneakernet Renaissance: Living Without the Cloud
Discover why air-gapping is going mainstream in 2026 and how organizations are securing local AI models using "digital vaults."
#44: When AI Trusts Too Much: The Art of Prompt Injection
AI's Wild West: Battling prompt injection and poisoning. Discover how AI threats are shifting from sci-fi to insidious attacks on the models...