Category: Governance
-
Documentation Is Now a Control, Not an Afterthought
The dangerous workflow is not always the one that fails. Sometimes it is the one that keeps working. This post explains why documentation has become a control for automated and agentic workflows, especially when ownership, scope, permissions, evidence, and operating intent drift over time.
-
Receipts or It Didn’t Happen: Evidence-Grade Logging for Agents
When agents read data, call tools, update records, and trigger downstream work, the final answer is not the whole story. Evidence-grade logging gives teams the connected receipts they need to explain what happened later.
-
Joiner-Mover-Leaver for AI Agents: Create, Rotate, and Retire Access Before It Drifts
AI agents need lifecycle management because they can hold credentials, inherit permissions, call tools, and act across systems. Learn how joiner-mover-leaver maps to create, rotate, and retire controls for agentic AI and non-human identities.
-
The Workflow Got Faster. The Record Got Fuzzier.
AI can make a workflow faster while making the audit trail less truthful. This post looks at how tool use, delegated authority, and fuzzy attribution collide once systems start acting across tools, why productivity is still the point, and what identity, logging, and governance controls have to catch up afterward.
-
RAG Is Data Access: Retrieval Authorization Is the Control
RAG is not just an AI retrieval pattern. In enterprise environments, it is a new data access path. This post explains why retrieval authorization is the real control, where the boundary actually sits, and what security, IAM, and platform teams should review before connecting sensitive content to assistant workflows safely.
-
Practical Retrieval Authorization Patterns for RAG Systems
Teams are adopting AI assistants fast, often without treating retrieval as a new access path. This practical guide explains how to secure RAG systems with real authorization patterns: preserve permissions through chunking, enforce policy at retrieval time, separate trust zones, and keep adoption moving with guardrails instead of friction intact.
-
Secrets and Tokens: Rotation SLAs, Blast Radius, and Attacker Dwell Time
Long-lived secrets are not harmless leftovers. They are blast radius waiting for an incident. This post breaks down rotation SLAs, attacker dwell time, delegated OAuth risk, and the operational habits that keep non-human credentials from becoming tomorrow’s breach path. Built for practitioners who have to run the mess in production.
-
Agents Are Identities: The New Control Plane for Enterprise AI
AI agents are not just features. Once they connect Slack, Jira, internal data, and downstream tools, they become operating identities with permissions, authority, and blast radius. This post explains why agent governance is an identity problem, not a prompt problem, and offers a framework: name, scope, bind, watch, prove, retire.
-
Agent Inventory and the Agent Register: The Control You Need Before Agent Sprawl Becomes Identity Debt
AI agents do not create a new governance problem so much as a faster identity problem. This post explains why agent inventories, Agent Registers, and ownership discipline matter, how discovery differs from authorization, and what teams need to prove scope, effective access, lifecycle, and accountability before something breaks in production.
-
Approved Tool, Expanding Agent: The Ownership Model That Works
Approved SaaS platforms keep gaining agentic capability. The original security review does not expand with them. This post covers the RACI model, re-review triggers, and anti-patterns that break accountability when trusted tools evolve, plus what NIST AI RMF and OWASP Agentic Top 10 say about why one-time approval was never…
Tech That MattRs 