Electronic health records and artificial intelligence systems complicate evidentiary assessment in medical criminal proceedings. This study identifies the conditions under which technical data can support proof of fault and attribution of injury to an identified person. Using doctrinal analysis and legal interpretation, it examines Iranian rules on electronic evidence, judicial knowledge, and medical liability, alongside functional comparisons with France, Germany, the United States, and technology governance instruments. The central finding is that admissibility, evidentiary weight, and sufficiency must be distinguished through five levels: provenance and authenticity; integrity and chain of custody; attribution and effective control; algorithmic auditability; and evidentiary sufficiency. A recorded warning or account activity does not independently establish awareness, fault, or causation. Liability for diya must also be distinguished from discretionary criminal punishment, and aggregate system performance from the validity of an individual output. Read alongside specific evidentiary rules, Article 120 of the Iranian Islamic Penal Code prevents material doubt from being replaced by presumed responsibility; it does not independently justify extinguishing patients’ financial rights. The study proposes a staged judicial reliability framework and a principle against inferential leaps from data to liability. Implementation requires proportionate evidence preservation, multidisciplinary expertise, meaningful opportunities to challenge evidence, and explicit reasoning in judicial decisions. The proposed framework is normative and requires independent empirical validation.