What the EU AI Act actually requires from your AI workflows
A practical read on what the EU AI Act asks of AI workflows: Article 12 automatic logging, Article 14 human oversight, and why append-only hash-linked evidence is the spine of compliance.
Notes on governing AI in production
Audit trails that hold up, deterministic replay, EU AI Act evidence and agent reliability engineering. Written by the people building Trazium — practical, sourced and without hype.
Topic
Deterministic replay for AI agents: turning "it failed" into "here's exactly why"
How deterministic replay turns an AI agent failure from a guessing game into a line you can point at, using event sourcing to record every model call, tool result and decision as immutable events.
Trazium vs n8n for regulated environments: an honest comparison
An honest comparison of Trazium and n8n for regulated environments: where general-purpose automation stops, and why append-only logs, deterministic replay and recorded oversight matter under audit.