Thoughts on AI agent security, data isolation, and why your AI doesn't need to know who you are.
How to give AI agents access to private data, expose them to untrusted content, and let them communicate externally — without the kill chain working.
The security model behind RedactSure: data isolation, user control, and policy enforcement for browser agents.
How architectural anonymity solves the insider threat that alignment never will.
What happens when we try to separate expertise from access? The answer is surprisingly asymmetric.