Kmori·Oct 3AI Made Finding Bugs Cheap. No Maturity Model Counts What That Costs.Every AI-SDLC maturity model measures how little the humans have to do. curl’s decision to shut down its bug bounty is a useful signal of…
Kmori·Aug 9A Decision Is Not a ControlYour policy engine said no. Did anything stop? The industry agreed that authorization belongs outside the model — the easy half. What makes…
Kmori·Aug 2Clean Was Never a Property. Four Fields Are Discovering That Separately.Multi-agent approval, institutional epistemology, and cryptographic attestation are arriving at one idea from three directions — that being…A response icon1A response icon1
InSystem WeaknessbyKmori·Jul 22The AI Finds the Flaw. Someone Still Has to Answer for the Fix.Everyone in security is asking the same question about AI — the young ask if the field is still worth entering, the rest of us ask what our…
Kmori·Jul 7The Unbreakable System Is the One That Breaks Correctly.Agent security is converging on containment — kill-switches, sandboxes, blast-radius limits. It’s the wrong instinct, and information…A response icon1A response icon1
Kmori·Jul 2OSS Never Solved Security. It Solved Responsibility.The open-source-versus-proprietary debate was always about responsibility, not security. The EU’s Cyber Resilience Act — with obligations…
InSystem WeaknessbyKmori·Jun 28We Spent a Century Killing Security by Obscurity. Agents Are Reviving It.The industry learned, the hard way, that hiding the system is not the same as securing it. Agent governance is quietly forgetting the…A response icon4A response icon4
Kmori·Jun 27Everyone Standardized the Inside. The Border Belongs to No One.2026 was the year agent governance went open-source and production-ready — for the agents you own. The interesting decisions happen where…A response icon5A response icon5
Kmori·Jun 15I Authorized the Agent. I Never Authorized What It Called Next.Single-agent authorization holds. Then the agent calls another agent, and the thing I built starts describing a world I’d already written…
Kmori·Jun 4The Threat Model Was Right. The Effort Was the Bug.Macro boundaries, micro assets, and the STRIDE-per-asset category error in mTLS-heavy environments