An LLM kept doing exactly what I told it, and that was the problem. Maybe the lesson isn't about fixing the AI but about being better communicators.
Working through something similar? I advise security and engineering leaders on building programs that get adopted rather than worked around.