ML safety
What ISO/PAS 8800 actually demands — a working checklist
The ML governance evidence we package for ISO/PAS 8800 today, and where it overlaps cleanly with the EU AI Act high-risk regime.
From the bench.
Working notes from active safety programs — what we're seeing in assessor rooms, what agents are getting right, and where the standards are catching up to physical AI.
Autonomy
SOTIF triggering conditions: stop arguing, start framing
A working pattern for scoping ODD, naming triggering conditions under ISO 21448, and getting residual-risk arguments through assessor review.
Functional safety
Why your FMEDA keeps getting bounced — and a faster fix
Three patterns we see in assessor rooms when ISO 26262 Part 5 FMEDAs come back marked up, and the agent prompts we use to land them in one round.