As soon as an AI collaborator starts logging multimodal memories about a user, what’s the first architectural safeguard you bolt on to guarantee those rich context traces stay helpful—yet purgeable on demand—to meet looming global privacy rules?