TMK

Research notes

Investigations and benchmark results

Each note is written from a completed research record: a fixed answer, the sources that were opened and read, and the searches that were run to try to break it.

Published field studies

Four short, source-led studies on AI and public power

Each one answers a narrow question from primary documents — regulation text, a court judgment, two public oversight audits and a federal enforcement record — with every claim mapped to the source it came from.

Election deepfake disclosureFacial recognition in public spacePredictive policing auditsAutomated hiring and discrimination

Published

AI assistants and Ewe health and public-service questions

Thirty-six recorded answers from three models to six household questions, asked in Ewe and in English, with all transcripts published. Fluent-speaker review is still outstanding, so no claim is made about accuracy or naturalness.

Read the published findingsRun-by-run tracker (6 models, 144 answers)How each benchmark is researched