Retention Clause Weighting

Retention Clause Weighting treats data-retention terms as a first-class selection criterion for foundation model platforms, ranked alongside price per million tokens and context length rather than after them. The logic: agencies rarely buy a model for themselves. They buy it on behalf of a client whose contract may specify data residency, deletion windows, or prohibitions on training use, and a platform that cannot meet those terms turns a working prototype into an unpublishable one. Weighting means scoring each candidate platform on retention posture before benchmarking quality, because quality is negotiable and a client's legal review is not. A practical example: an agency scoping a retrieval pipeline for a regulated client should confirm private deployment options before committing engineering time, since Cohere offers private VPC and on-premises deployment while many pay-per-token APIs do not. Forrester's warning that vendors simultaneously set usage limits, control feature access, and arbitrate their own disputes makes retention terms a governance question, not a footnote.