Maestro University Chatbot Exposes Evaluation Criteria

This is a transparency bug dressed up as helpfulness: once a decision system explains its rubric, it also teaches users how to game it.

• –DeepMind's metacognition track is a real fit for the claim, but the exploit itself is still an anecdote until someone else reproduces it.
• –Education, admissions, hiring, and triage workflows are all exposed when the bot reveals the exact signals driving ranking or priority.
• –The core fix is disclosure control, not just better prompting: separate user help from decision logic and clamp what the system can reveal about itself.
• –MBSE is a sharp benchmark concept, but the stronger contribution would be a reproducible test suite with clear baselines.