Why enterprises need bounded, observable, and governed improvement loops for production AI agents, and how the open-source ASSERT framework engineers the half of the loop everything else depends on.
The improvement loop is also the record a supervisor will eventually ask to see. Under the AI Act as amended by the AI Omnibus, in force since 27 July 2026, the Annex III high-risk rules apply from 2 December 2027 and Annex I embedded products from 2 August 2028.
That regime wants something close to your artifact list: a written behavioural contract, evidence the system was tested against it, and a named human at the gate. Teams already versioning taxonomies and verdicts will be assembling a file rather than inventing one. Whether supervisors accept judge verdicts as evidence is still an open point.
The improvement loop is also the record a supervisor will eventually ask to see. Under the AI Act as amended by the AI Omnibus, in force since 27 July 2026, the Annex III high-risk rules apply from 2 December 2027 and Annex I embedded products from 2 August 2028.
That regime wants something close to your artifact list: a written behavioural contract, evidence the system was tested against it, and a named human at the gate. Teams already versioning taxonomies and verdicts will be assembling a file rather than inventing one. Whether supervisors accept judge verdicts as evidence is still an open point.