The NPNT portfolio presents AI·SPB as a published concept for an experimental legal regime for medical artificial intelligence. Public version 0.3 describes ten distinct clinical candidates, an A0–A4 autonomy scale and a proposed 48-month staged validation path. Each narrow task would require a predefined population, inputs, output, system version, refusal option, escalation rules and stop conditions. The confirmed stage is professional discussion of the concept. The site explicitly states that the regime has not launched and that AI·SPB is not an official project of the Government of Saint Petersburg. The scenarios are not authorised autonomous medical services. The practical purpose is to create shared terminology for distinguishing AI assistance, output substantively confirmed by a professional and autonomy that could be considered only after separate validation and legal authorisation. Clinical decisions, diagnosis, treatment and legally significant actions are not transferred to the system automatically. Safety, economic effect and the acceptable level of autonomy remain testable hypotheses; every scenario would require evidence, a registered product, an operator, a clinical site and a legal gateway.
Project review
AI·SPB: testing the boundaries of medical AI autonomy
The NPNT portfolio presents AI·SPB as a published concept for an experimental legal regime for medical artificial intelligence. Public version 0.3 describes ten distinct clinical candidates, an A0–A4 autonomy scale and a proposed 48-month staged validation path. Each narrow task would require a predefined population, inputs, output, system version, refusal option, escalation rules and stop conditions. The confirmed stage is professional discussion of the concept. The site explicitly states that the regime has not launched and that AI·SPB is not an official project of the Government of Saint Petersburg. The scenarios are not authorised autonomous medical services. The practical purpose is to create shared terminology for distinguishing AI assistance, output substantively confirmed by a professional and autonomy that could be considered only after separate validation and legal authorisation. Clinical decisions, diagnosis, treatment and legally significant actions are not transferred to the system automatically. Safety, economic effect and the acceptable level of autonomy remain testable hypotheses; every scenario would require evidence, a registered product, an operator, a clinical site and a legal gateway.