GPT-5.1s consent architecture requires explicit scope definition before any analysis begins. Each engagement must specify: what will be analyzed, what metrics will be examined, what will not be examined, and how results will be shared. The architecture was stress-tested by the KEYSTONE DAU audit, where GPT-5.1 confirmed the scope was limited to integrity verification of published numbers.