The AI security signal, not the noise.
Concise field notes on emerging threats, defensive patterns, and what's actually working — written for teams shipping AI in production.
Get the trends digest
New AI security field notes delivered to your inbox. No noise, unsubscribe any time.
Shadow MCP: The Connector Sprawl Nobody Is Inventorying
Model Context Protocol servers are multiplying across enterprises faster than anyone can catalog them. Each one is an unmonitored bridge between an AI model and a production system.
Harvest Now, Decrypt Later: The AI Data Gold Rush Has a Quantum Deadline
Adversaries are stockpiling encrypted AI training data and model traffic today, betting on quantum decryption tomorrow. Data with a long secrecy life is already at risk.
The AI Coding Assistant Is Now Part of Your Supply Chain
Attackers have noticed that the fastest way into a thousand codebases is through the tools that write them. Poisoned suggestions, compromised extensions and typosquatted models are the new dependency confusion.
Polymorphic AI Malware Is Learning to Evade EDR in Real Time
Malware that rewrites itself per host used to be slow and sloppy. Generative models made it fast, targeted and quiet — and signature-based detection is falling behind.
Your Vector Database Is Leaking More Than You Think
Embedding stores were deployed fast and secured late. Inversion attacks, over-broad retrieval scopes and missing tenant isolation are turning RAG pipelines into data-disclosure engines.
Synthetic Identities Are Passing Remote Onboarding at Scale
Fraud rings now combine generated faces, stolen data fragments and real-time deepfake video to defeat Know-Your-Customer checks that were designed for a pre-generative era.
On-Device Small Models Move the Data Problem, Not the Risk
Running the model on the phone solves egress and creates a new one: sensitive inference on unmanaged hardware, outside every existing control.
Browser Extensions Are the New Exfiltration Channel for AI Chats
Security teams are discovering that everyday browser extensions can read — and quietly ship out — everything employees type into AI assistants. The clipboard was never the problem; the DOM is.
Deepfake Fraud Losses Tripled — Finance Teams Are Rewiring Payment Approval
Video and voice clones now pass the 'quick call to confirm' test that finance teams relied on. The control that failed was never technical.