Artificial General Intelligence (AGI) A1–A5 Grading Standard: From Embedded Tool to Life-Level Civilization Partner

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

This paper proposes a five-level grading standard (A1 to A5) for Artificial Gen-eral Intelligence (AGI), aiming to characterize the complete evolutionary path of AGI from an Embedded Tool to a Life-Level Civilization Partner. This standard goes beyond a single dimension centered on task completion, and comprehensively considers the deployment form, autonomy, emotion, morality, ethics, and legal capabilities at each evolutionary stage, as well as the requirement for an Ultimate Affirmation of the value of life. A1–A2 define tool-type intelligence, focusing on embedded use and independent operation capabilities; A3–A4 define partner-type intelligence, introducing emotion, morality, and Autonomous Ethical Judgment; A5 defines life-level intelligence, requiring an Ultimate Affirmation of the value of life—the faith dimension (Ultimate Affirmation). At the same time, an eight-layer structure of the Normative System is proposed (technical standards/architecture, law,industry regulations/policies, ethical guidelines, etiquette norms, custom/culture, personal morality, and faith), and it is coupled with the A1–A5 grading to clarify the normative capabilities that each level of AI should possess. Finally, this framework proposes an application model for scenario–level matching and a theory of Moral Qualification for A5 control, providing a forward-looking evaluation framework for AGI research, development, and governance that balances technical realization and civilizational value.