Curated signal, practical education, live demonstrations, and trusted guidance for individuals, teams, high-stakes and cleared organizations.
AI Agents are a security nightmare. They catastrophically walk face first into the original sin of computing: mixing instructions with data. Also, if you read on, you'll learn what prompt injection is. So, read!
The fix for prompt injection already exists — we built it once, in the telephone network, and then destroyed it ourselves for commercial reasons. It's called channel separation. This article explains what we had, what we lost, what it would take to rebuild it for AI, and why we need to do it yesterday.
A recent peer-reviewed paper from UC Berkeley (StruQ, Chen, et al, USENIX Security 2025) successfully implements a separation of control and data channels for LLMs, effectively reducing prompt injection success to near zero while maintaining utility. However, the solution applies to programmatic single-turn API calls, leaving the hard problem—recursive multi-turn conversational agents—unresolved. The paper proves the structural separation principle works, but underscores the need for a standardized protocol over individual vendor products.
This free course is for anyone using AI tools without a clear framework for the security and privacy risks that come with it.
If you've heard about AI but haven't developed a systematic understanding of what it actually is or how to use it, this course is for you.
If you've decided to bet on yourself — whether by choice or circumstance — you know that running a business alone requires you to do the work of multiple people.
You've learned to use AI, but you're still doing the manual glue work — copying, pasting, switching between tools, running the same prompts every week. This course closes that gap.
This course is for managers, team leads, and department heads who've already learned to use AI themselves and now face a different problem: getting their entire team to work differently.
This course is for technically curious people who have outgrown traditional AI tools and want to build something that works for them while they're doing something else.
Running visualization of a spiking neural network demonstrator that you can interact with.
AI literacy, advisory, and adoption guidance from people who have operated in classified environments — for organizations where getting it wrong is not an option.
Checking auth state...