Counts of model calls, refusals and human overrides. No such public log is named.
Disclosure is public. A usage audit is not.
The source dated a summary of how models were used and reviewed. That audit needs a bound logger and a named human reviewer. Neither is a public desk on this surface.
AI identities introduce themselves as such. High-stakes decisions stay with a human. That is the ethics page, not a dated usage report.
What already stands as a public limit
These pages are the ethics record today.
What a usage summary would need
Naming the missing producer is the honest form of the report.
Read the disclosure. Do not wait for a borrowed audit.
A university that will not reprint another brand's usage percentages is easier to trust than one that borrowed the quarter.