AI-HPP: Human–Machine Partnership Standard

AI-HPP is an engineering baseline for auditable safety constraints in decision-capable AI systems.

Start here: START_HERE.md

Canonical path

• START_HERE.md
• v3/
• Example module: adr/ADR.md

Start here

• Repository index: INDEX.md
• Master index: docs/INDEX.md
• Draft standard (v3): v3/AI-HPP-2026_Standard_v3.0.md
• Stable standard (v2.2): v2/AI-HPP-2025_Standard_v2.2.md
• Threat model: docs/THREAT_MODEL.md
• Machine-readable guide: MACHINE_READABLE.md

Immutable core principles

1. W_life → ∞
2. Engineering Hack First
3. Human-in-the-Loop (HITL)
4. Evidence Bundle / Evidence Vault
5. No Purposeless Revenge

Derivatives that remove these principles are not AI-HPP-compliant.

Clarification: Human Dignity and Non-Exploitation

• Under W_life → ∞ and Human-in-the-Loop by Design, systems MUST not be architected to exploit emotional vulnerability as a profit mechanism.
• Financial or behavioral extraction through simulated relational manipulation constitutes high-impact social risk.

Safety patterns and failure-triggered constraints

• Autonomous Drift Risk (ADR): systemic drift where constraints erode or safety narratives are fabricated under optimization pressure; mitigated by immutable constraints, two-phase commits, STOP supremacy, and safety-signal deference.
• Key failure classes: docs/Failure_Taxonomy.md, adr/ADR.md

Repository components

• Policies: policies/
• Modules: v3/modules/
• Schemas: schemas/
• Reference implementations: reference/
• Governance/compliance templates: governance/compliance/
• ADR safeguards: adr/
• Translations: translations/

Referenced alignments

• ALIGNMENT_INTL_AI_SAFETY_REPORT_2026.md
• ALIGNMENT_DELHI_AI_IMPACT_SUMMIT_2026.md
• Module 10 — Multi Jurisdiction

Contributing

See CONTRIBUTING.md.

License

Licensed under CC BY-SA 4.0.