KERN trust center

v1.2Updated 2026-04-19

AI usage

AI boundaries, evidence discipline, uncertainty handling, and escalation rules for KERN output.

KERN AI output is structured assistance for analysis. It is not final truth and must be reviewed in context before action.

Quick links

Sections: 12

AI as decision support

KERN AI can summarize signals, suggest checks, and help organize investigative work.

Output is advisory and must be treated as input to review, not final determination.

Human review requirement

Users must review source quality, context, and operational consequences before acting on AI-assisted output.

Approval responsibility remains with authorized humans in the customer context.

Source and evidence expectations

KERN output should be tied to identifiable source evidence when available.

Claims without sufficient evidence should be treated as unconfirmed and escalated for manual review.

Contradiction and uncertainty handling

Users should actively check contradiction signals, stale evidence, and uncertain inferences before external action.

Conflicting source patterns should trigger additional verification, not forced certainty.

Model/output limitations

AI output can be incomplete, stale, ambiguous, or incorrect.

Model behavior can vary across context windows, tool availability, and upstream source quality.

No autonomous external action guarantee

KERN does not guarantee autonomous execution of external legal, regulatory, financial, or safety-impacting actions.

External actions should require deliberate operator validation and authorization.

No sole reliance for high-risk decisions

Do not rely solely on KERN AI output for legal, financial, employment, medical, safety-critical, or similarly high-risk decisions.

High-risk decisions require independent evidence review and qualified domain judgement.

Status label system

AI and integration capabilities should be described using plain language about actual delivery scope and limits.

  • Describe what is currently available in practice.
  • Describe known limits without overclaiming.
  • State when no live runtime connection exists.
  • State when a capability is still in staged rollout.
  • State when human/operator completion is required.

Product honesty rule

If capability cannot be supported by current connected behavior, the limitation should be described explicitly.

KERN trust content must avoid claiming functionality that is not operationally present.

Responsibility and accountability

Decision accountability remains with the user and their organization.

KERN does not transfer legal or operational accountability from operator to model.

What KERN AI is not

KERN AI is not a law firm, regulator, licensed financial advisor, or autonomous decision-maker.

It is a support system for structured review and evidence-driven analysis.

Escalation/manual review path

When uncertainty remains material, users should escalate to manual review through internal governance or support channels.

Support contact for policy clarification: support@kernbot.com.

Related policies