https://mhl42.ai/blog Notes from the Mostly Harmless practice on securing AI agents. en https://mhl42.ai/blog/trace-for-agentic-ai https://mhl42.ai/blog/trace-for-agentic-ai Tue, 01 Sep 2026 09:00:00 +0200 Version 0.1 of the agentic extension is public. It keeps every object, phase and gate of the original TRACE and adds an applicability screen, agentic model objects, six optional phases, and crosswalks to MITRE ATLAS and the NIST AI RMF. https://mhl42.ai/blog/trace-and-human-ai-co-working https://mhl42.ai/blog/trace-and-human-ai-co-working Tue, 25 Aug 2026 09:00:00 +0200 AI-human co-working was one of the three founding drivers of TRACE. That design decision is why the method holds up when the system being modelled is itself an AI agent. https://mhl42.ai/blog/zero-trust-for-agents https://mhl42.ai/blog/zero-trust-for-agents Wed, 12 Aug 2026 09:00:00 +0200 Anthropic's May 2026 guide applies zero trust to agent deployments in three tiers and eight phases. What it gets right, where the tiers are off, and three things to add.