MAF paragraphs 3.55 to 3.59: test user interfaces for usability before deployment and tell users when their responses train the system; communicate in easy-to-understand language, measuring readability and using visual aids for higher-impact decisions; for systems that learn from live input, set acceptable user policies barring manipulation, reverse engineering or disruption; and obtain from AI solution providers the information and features (for example data lineage and user preference controls) needed to follow the framework, including by back-to-back arrangements.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.