How is Burna AI different from general-purpose clinical AI tools?

Burna AI does one job, adverse event grading and attribution, and constrains it by architecture. Specificity: it grades CTCAE adverse events, codes them to MedDRA and scores attribution per drug; it does not diagnose, recommend treatment or offer general decision support. Transparency: every grade traces to the line in the medical record that produced it and to the CTCAE criterion it meets, and the WHO-UMC and Kramer reasoning behind each attribution score is shown. Discipline: twelve specialized agents run in a cascading constraint pipeline, so each agent's output bounds what every later agent can return; a grade must be a valid CTCAE grade, and a grade with an incomplete citation set is flagged before it is signed. An investigator signs every result. AI suggests, clinicians decide.