Clear boundaries for serious AI work.
Current policies, product-processing disclosures, reliability limits, account controls, and support paths—without turning legal language into marketing proof.
Trust starts with saying exactly what the system does—and what it does not.
Coherence uses a separate adjudication and repair boundary to reduce AI reliability risk. Risk reduction is not a guarantee, and consequential decisions still require the appropriate human authority.
The policy or control path you need.
Each reader page below is synchronized to the current product policy and links through to the product-domain canonical copy.
Data and privacy
Data categories, purposes, storage, retention, provider processing, and available choices.
Read Privacy →Security & Trust
Identity, data handling, separated authority, application integrity, subscriptions, lifecycle controls, and responsible disclosure.
Review the security posture →System specification
The public contracts for assertion release, composite repair, bounded orchestration, context continuity, replay, and replaceable resources.
Read the specification →AI provider processing
How prompts and relevant context may cross service boundaries for a selected live feature.
Review processing →Reliability boundaries
What CRE controls, where source quality still matters, and what remains outside the guarantee boundary.
Review limitations →Terms of Service
Reliability boundaries, user responsibilities, provider processing, subscriptions, and acceptable use.
Read Terms →