Safety & Trust
Know what is — and is not — being claimed.
This page sets a simple boundary: research and local previews are not released product behavior. Memory, developer-platform, availability, security, and compliance claims stay inactive until the exact customer-facing release proves them.
Stage labels matter
A research direction describes a question. A local preview demonstrates layout. Neither is a live feature, product offer, or availability claim.
No active continuity claim
LongMind is a research direction. It does not describe a current ChatNova capability.
No active platform claim
The Console is a local sample-data preview. It does not provide developer credentials, endpoints, models, tiers, pricing, billing, or platform access.
Proof before release language
A later claim needs the exact released artifact and a reproducible customer-facing result, including the failure and control cases that matter to the claim.
The system
What the labels mean
Research describes a question or proposed design. A local preview can show sample data and interaction structure. Released behavior requires an exact customer-facing artifact that reproduces the claim. These states are not interchangeable.
What a later continuity release must prove
A later continuity release must provide clear user choices and prove their effect across real product use. Until that customer-facing evidence exists, no continuity or privacy guarantee is active here.
What a later platform release must prove
A developer-platform release must prove real credential lifecycle, authenticated requests, release identity, usage and cost controls, current documentation, support, rollback, and an end-to-end workload. A model claim additionally needs reproducible evaluation at the exact released boundary. Sample Console data proves none of those things.
What a later trust claim must prove
Security, privacy, residency, and compliance language must match the exact deployed data flow, adopted policies, operating controls, support process, and any required independent certification. Architecture intent, code, or a roadmap cannot activate those customer-facing claims.
What this page says now
Research and local previews are labeled as such. They do not activate memory, model, platform, availability, security, privacy, residency, or compliance claims.
What changes the boundary
Only a released artifact, customer-facing end-to-end evidence, relevant negative controls, and public terms that agree with runtime behavior can promote a claim. The linked research remains research until that happens.
The boundary is plain: proposal is not release. Read the research.