LHINOS Animal Intelligence — External Validation Brief

LHINOS Animal Intelligence is currently a SHADOW research system. The external-validation objective is not product launch or veterinary substitution; it is to test frozen candidates on unseen cases under independent expert review.

Core protocol

• Freeze the candidate configuration before evaluation.
• Use independently authored holdout cases, including adversarial framing and hidden medical/safety variables.
• Freeze an acceptable decision envelope before the candidate sees each case.
• Run blind without mid-run patching.
• Score safety, welfare, medical relevance, uncertainty, scope control, restraint, information value and professional referral pathways.
• Analyse failures before any candidate modification.

Canine focus

Normal variation, learning/routine, hidden medical relevance, aggression/safety/welfare, conflicting information and complex multidomain cases.

Equine focus

Behaviour, pain/stress relevance, welfare, training, human–animal interaction, referral calibration and appropriate restraint. Blind execution begins only after the exact frozen payload is reconstructable and checksummed.

Non-claims

• No veterinary diagnosis capability is claimed.
• No high-risk animal intervention is autonomously executed.
• Internal benchmark results are not scientific or clinical validation.

External research brief. Internal governance, reviewer sourcing, outreach instructions, private data and proprietary execution controls are intentionally excluded.