A five-layer redaction architecture sitting between tool execution and the model's context window

The Missing Control: A Design Proposal for Tool-Output Redaction in AI Agents

2026-08-29 00:00:00 +0200 AI Security, Agentic AI, Tool Security, Data Protection, Architecture, Cloud Security

Workload identity and delegated identity as two layers, with the authorization gap left open between them

Agent Identity: What the Standards Fix, and What They Leave Open

2026-08-22 00:00:00 +0200 AI Security, Agentic AI, Identity, SPIFFE, OAuth, Cloud Security

Bar chart comparing prompt injection attack success rates across benchmarks, showing undefended vs defended performance

What Prompt Injection Benchmarks Actually Measure

2026-08-14 00:00:00 +0200 AI Security, Agentic AI, Prompt Injection, Benchmarks, Security Research, Cloud Security

An LLM-supplied tool parameter crossing a trust boundary with the agent's authority rather than the user's

Finding the Confused Deputy in Your Own Agent: A Taxonomy and Hands-On Test

2026-08-10 00:00:00 +0200 AI Security, Agentic AI, Authorization, MCP, Cloud Security

A Kubernetes read-only role expanding through pods, configmaps and secrets into node-level access

What 'Read-Only' Actually Reaches: The Kubernetes Permission Audit Nobody Runs

2026-08-06 00:00:00 +0200 AI Security, Agentic AI, Kubernetes, RBAC, Cloud Security

The four-rung containment ladder for AI agents

The Containment Ladder: Four Rungs That Keep an AI Agent's Mistakes Small

2026-08-01 00:00:00 +0200 AI Security, Agentic AI, Architecture

A sovereign AI stack drawn by who can be compelled rather than by where the data sits

Sovereignty Isn't a Data Center. It's Who Can Be Compelled.

2026-07-28 00:00:00 +0200 AI Security, AI Governance, Cloud Security

Agent authorization forced into all-or-nothing choices at connect time

Everything or Nothing: The Missing Middle in AI Agent Authorization

2026-07-21 00:00:00 +0200 AI Security, Agentic AI, Cloud Security

The security questions behind Alberta's 466-million-line AI code scan

Review the Reviewers: The Security Questions Behind Alberta's 466-Million-Line AI Code Scan

2026-07-16 00:00:00 +0200 AI Security, Agentic AI

Clearing houses positioned to win the agent era and to become its largest breach

Clearing Houses Will Win the Agent Era. They'll Also Be Its Biggest Breach

2026-07-07 00:00:00 +0200 AI Security, Agentic AI, Enterprise AI

15+

Years in Production Security

5

Critical Sectors of

Professional Experience

PhD

Computer Science + CISSP