Governance

Policy-bounded autonomy. Human oversight on demand.

SOHMA operates within explicit operator-defined boundaries. Low-risk adaptations are automated; higher-risk actions surface for human review.

Current operating envelope across all surfaces

Low-risk adaptations

Tone, pacing, clarification

Continuously applied. Fully logged and auditable.

Automated within policy

Moderate-risk adaptations

Reflection, source confidence

Observed in production; sampled for human review.

Monitoring

High-risk adaptations

Strategy change, content gating

Flagged for operator review before broad rollout.

Review suggested

Human-required actions

Escalation, safeguarding, refusal

Routed to a human reviewer by policy, every time.

Escalation required

Distribution of adaptations by governance mode

Automated within policy71%

Monitoring (no action)18%

Review suggested8%

Human escalation3%

Humans do not approve every micro-adaptation. Operator policy defines what runs automatically, what is monitored, and what must escalate.

Policy version

v2.4.1

Audit log retention

365 days

Decision traceability

100%

Human reviewers on-call

24 / 7