Blog
Research and engineering notes.
An ongoing series on delegated authority and least-privilege enforcement in multi-agent AI systems. We publish the reasoning rather than the conclusions, because the reasoning is the part you can check.
Latest article
CERT-In’s AI Agent Advisory (CIAD-2026-0049): What Happened, What It Requires, and What to Fix First
On 15 October 2026, CERT-In issued a critical advisory on AI agents as attack vectors. The weaknesses it describes are old ones: exposed credentials, forgotten APIs, over-permissive roles. What is new is the speed, and the fact that not one affected organisation noticed.
Read the article →More articles
Stop Blaming Your Security Team for AI Agent Breaches
The real problem is not effort. Nobody could have set the right permissions, because the system does not allow it.
Read →The Access Control Gap in AI-Native Architectures
Conventional IAM was architected for human users at predictable entry points. Multi-agent systems are a different problem, not an extension of the same one.
Read →The research behind these notes
Five findings on AI agent authorization
A systematic review of 29 primary works, each finding with its evidence and its sources.
Read the research →Also on LinkedIn