Responsibility
Named humans remain accountable for consequential decisions.
Trust
Trust is a system property, not a promise made after the fact.
Accountability, consent, evidence, audit, action controls and model governance are designed into the platform.
Human accountability
Responsibility stays with people
Responsibility, authority, delegation and escalation are explicit and recorded. The system cannot create its own permission; high-impact action remains policy- and authority-bound.
Authority & delegation
Consent, access, delegation and permission to act are separated and recorded.
Escalation
When the scene is not ready, the system clarifies, constrains or escalates.
Consent & protected perspectives
Purpose limitation and disclosure controls
Purpose limitation
Information is used for its authorised purpose, not beyond it.
Consent & revocation
Consent can be given and revoked; disclosure controls govern what is shown.
Protected perspectives
"Some relevant context is protected and is not shown in this view."
Evidence & audit
Evidence chains and decision audit
Every consequential decision and action is evidence-linked and audited. Source provenance matters because it is what makes memory challengeable.
Evidence chain
Assertions stay linked to their source evidence and runtime history.
Activity & decision audit
Who decided, when, on what evidence, and with what alternatives.
Source provenance
Why source provenance matters: it is what allows challenge and correction.
Knowledge integrity
Contradiction, confidence, time and correction
Contradiction
Conflicting memories coexist explicitly rather than being silently merged.
Confidence & temporal validity
Uncertainty and the time something was true are explicit states.
Supersession & correction
An assertion can be corrected without erasing the original source evidence.
Action controls
Useful automation with accountable boundaries
Policy & quality checks
Output and action are validated against policy, authority, evidence and quality controls.
Approvals
Proposed actions require the right role or approval before execution.
Idempotency, rollback & audit
Authorised actions are idempotent, reversible where possible and audited.
Model governance
Provider-neutral, governed and observable
Provider-neutral contracts
Identity, decisions and scene contracts are provider-neutral.
Promotion & rollback
Behavioural compatibility, identity stability and rollback readiness are checked before promotion.
Data handling & observability
Data handling, residency and observability are governed per approved policy.
Privacy lifecycle
Scope, retention, deletion and index propagation
Scope
Access is scoped by tenant and role.
Retention & legal hold
Retention rules and legal hold (where applicable) are respected.
Deletion propagation
Deleted data disappears from active retrieval views and derived indexes.
Claim discipline. We do not claim named certifications, legal compliance, encryption details, data-residency locations, uptime or security controls unless confirmed by the relevant owner.
Review trust by design with your team.
Talk to Humans+ about governance, privacy, evidence and accountability.