Trust
Safety includes human authority.
Merism should assist with difficult work without making consequential decisions on behalf of people.
Safety belongs throughout the work.
Explore six responsibilities. This model explains our approach; it does not certify a deployed Merism capability.
Start with the task and the boundary.
Identify the intended outcome, sensitive data, consequential actions, and conditions under which the system must stop.
A proposed workflow names its permitted inputs and tools before it begins.
Boundaries
A model may analyze, draft, or recommend. Permission to act requires an appropriate user, scope, and review step.
Known limitations
AI output can be incomplete or wrong. Users should check important facts and results, especially before consequential decisions.
Abuse
We are preparing monitored abuse reporting and a policy for unacceptable use before broader access.