Patent offices around the world have converged on one answer to a headline question and split on the question that actually governs practice. Everyone agrees an inventor must be a human being, so an AI system cannot be named. But how much a person must contribute when a machine does much of the creative work is contested, and the answer changes real filings. This UpLaw analysis maps the state of play across the United States, the European Patent Office, the United Kingdom, and the major Asian offices, anchored by the DABUS litigation that forced the issue and by fresh worked hypotheticals a founder, a judge, and a patent lawyer can all follow. It works through the conception doctrine at the core of the debate, the twin problem of disclosing a black-box invention under written-description and enablement law, and the practical steps that convert AI-assisted research into enforceable patents. Throughout, it treats inventorship not as a name to fill in but as a latent defect or a quiet strength, decided long before anyone reads a claim.