The ADAFP AI Ethics Charter sets out the rules the Laboratory works to. This page records where those rules are met and where they are not. It is updated as things change, which is why it lives here rather than inside the Charter — the Charter is adopted by the Board and changes rarely; this record changes often.
Publishing the gaps is the point. An organisation that names its own failings makes a claim that can be tested against it later, and that is the only kind of claim worth making about a political alliance operating tools of this sort. Nothing here is softened because it is inconvenient.
Met — the commitment is implemented and can be checked from outside. Partly met — implemented in part, with a stated shortfall. Open — not yet done. No commitment is marked met on the strength of an intention.
If you believe an entry on this page is wrong — including one marked met — say so at office@adafp.org. Under Charter section 4.5 you will receive a substantive reply within fourteen days.
Several rows can be verified without trusting us. The externally observable checks are automated in an audit agent that anyone may run against this platform; it uses no credentials, changes nothing, and reports what it requested and what came back. The functional questions this page cannot answer — whether outputs are accurate, whether citations are real, whether governments and oppositions are treated by the same standard — require an evaluator who did not build the platform.