Skip to content
LEGAL · RESPONSIBLE USE

Responsible Use Statement

KRYOS frameworks operate on enterprise data that frequently involves personal information, third-party rights, regulated content, and material business consequences. The principles below govern how those frameworks are designed to be used.

1. Rights-aware by default

No asset is recognised, valued, productized, or capitalized without an evidenced rights posture: lawful source, lawful purpose, lawful onward use, jurisdictional applicability, and counterparty contract review. Where rights cannot be evidenced, the asset is routed to repair, withhold, or exclude.

2. Provenance-linked evidence

Every recognition and measurement decision is tied to evidence with lineage that can be independently inspected. Outputs are designed for audit readiness; they do not assert audit conclusions.

3. Fail-closed recognition

Recognition gates default to do not recognize. Borderline cases are not silently passed through. The standard of proof is explicit, documented, and reviewable.

4. Current value separated from optionality

Today's defensible value is reported separately from hypothetical fusion upside, scenario tails, or AI-conditional optionality. Conflation is a misuse of the framework.

5. Subject to evidence and review

All KRYOS outputs are subject to independent review by appropriately qualified professionals - accounting, legal, tax, audit, valuation, actuarial, and regulatory - before they are relied upon for material decisions.

6. Personal and sensitive data

KRYOS does not encourage the collection of additional personal data and applies data-minimisation as a default. Sensitive categories (health, biometric, financial identifiers, children's data, etc.) are handled only where lawful, necessary, and evidenced under an executed agreement.

7. AI and model-related data

Training corpora, RAG indexes, embeddings, evaluation sets, and model telemetry are recognised only where the rights to use the underlying data for training and inference are evidenced. Synthetic data is treated according to its derivation chain, not by label.

8. Red-team challenge

Material outputs are designed to be tested by independent red-team review under the GAER layer. Outputs that have not been challenged should be treated as preliminary.

9. Honest representation

KRYOS outputs must not be presented externally in ways that overstate what they assert. Bands are bands. Illustrative is illustrative. Pre-engagement is pre-engagement.

10. Acceptable-use boundaries

KRYOS will not knowingly support engagements designed to: mislead investors, regulators, insurers, lenders, auditors, or counterparties; evade data-protection law or sanctions; recognise rights that the client does not hold; or productize data in violation of consent, contract, or applicable law.

See also: Disclaimer · Terms of Use · Privacy Notice.