AI trust-stack applicant concentration: verification and model-coding artifact

Public, network-free verification artifact for Who Owns the AI Trust Stack? Patent-Applicant Concentration in Standards-Dependent Infrastructure. It reproduces the eleven-field results for all eight reported concentration specifications. Conditional on supplied total and observed unit counts, it independently recalculates nine derived statistics from anonymous ranked fractional weights. It also contains title-free outputs from two free local language-model coding passes: Qwen 2.5 7B on 451 sampled records and Mistral 7B on the prespecified 80-record reliability subset. These are model decisions, not human coding. The release excludes patent titles, identifiers, applicant names, and raw EPO OPS or WIPO PATENTSCOPE-derived records. It reproduces the reported concentration calculations but does not rebuild the upstream 151,890-record patent corpus.