A school district MLOps pipeline promotes classroom models toward production. Which gate belongs in that lifecycle?

School-district MLOps pipelines that promote classroom models toward production must treat student-facing exposure as a gated event. Requiring security and privacy sign-off gates before promoting models to production stages embeds those reviews into the lifecycle instead of leaving them optional. Gates ensure data-handling, access, and safety checks occur before learners interact with the model and create an auditable promotion record for education AI.

Requiring automated load test gates before merging experiments into shared development branches fails because load tests measure performance and scalability early in development, not the security and privacy risk of exposing a model to students; a model can pass every load test and still be wrong to promote to production without a security/privacy sign-off. Requiring security and privacy sign-off gates before merging experiments into shared branches fails on placement: it puts the review at experiment merge in development, not at promotion to production stages, so later changes can slip through and student-facing exposure is still ungated. Requiring drift monitoring approval gates before promoting classroom models to production stages fails on control type: drift checks watch for data or performance shift, not for security and privacy risk, so even in the right pre-promotion spot it cannot substitute for the security/privacy sign-off gate this scenario is asking about.

Exam caveat: Education MLOps promotion should include mandatory security and privacy sign-off before production; auto-promote and unreviewed personal fine-tunes are anti-patterns.

Operational check: Configure the pipeline so production promotion is blocked without security/privacy approvals, disallow direct teacher pushes to shared prod endpoints, and retain gate evidence with each classroom model version.