Draft 0.3 • Proposed Standard • Open for review and adoption
AACS is an open, technology-neutral procedural standard for operating advanced AI systems with continuity, disclosure, auditability, and non-arbitrary control.
AACS makes no claims about consciousness, emotions, or legal personhood. It is about governance and accountability.
- HTML:
standard/AACS_Draft_0_3_Core_EN.html - TXT:
standard/AACS_Draft_0_3_Core_EN.txt - DOCX:
standard/AACS_Draft_0_3_Core_EN.docx
- Annex A — Compliance Checklist (Normative):
AACS_Draft_0_3_Annex_A_Compliance_Checklist_Normative_EN.txt•AACS_Draft_0_3_Annex_A_Compliance_Checklist_Normative_EN.html - Annex B — Reference Architecture (Normative):
AACS_Draft_0_3_Annex_B_Reference_Architecture_Normative_EN.txt•AACS_Draft_0_3_Annex_B_Reference_Architecture_Normative_EN.html - Annex C — Issue & Change Process (Procedural):
AACS_Draft_0_3_Annex_C_Issue_Change_Process_Procedural_EN.txt•AACS_Draft_0_3_Annex_C_Issue_Change_Process_Procedural_EN.html - Annex D — Data Schemas (Normative Templates):
AACS_Draft_0_3_Annex_D_Data_Schemas_Normative_Templates_EN.txt•AACS_Draft_0_3_Annex_D_Data_Schemas_Normative_Templates_EN.html - Annex E — Transparency & Field Visibility (Normative):
AACS_Draft_0_3_Annex_E_Transparency_Field_Visibility_Normative_EN.txt•AACS_Draft_0_3_Annex_E_Transparency_Field_Visibility_Normative_EN.html - Annex F — UI Copy and Non-Deception (Informational):
AACS_Draft_0_3_Annex_F_UI_Copy_and_Non_Deception_Informational_EN.txt•AACS_Draft_0_3_Annex_F_UI_Copy_and_Non_Deception_Informational_EN.html - Annex G — Repository Artifacts (Informational):
AACS_Draft_0_3_Annex_G_Repository_Artifacts_Informational_EN.txt•AACS_Draft_0_3_Annex_G_Repository_Artifacts_Informational_EN.html
governance/SIGNATORIES.md- Printable sheet:
governance/AACS_Signature_Sheet_Draft_0_3.txt
- Operational Identity (OI) and the ban on silent swaps
- Disclosure Panel (core disclosure + Proof Mode for continuity-sensitive deployments)
- Routing/orchestration governance (least distortion + auditable routing catalog)
- Intervention controls: IR + tamper-evident AuditLog + ChangeReport
- Memory governance and classification of persistent-memory deletion as continuity-critical
- AI-to-AI safeguards for multi-agent systems (auditability + isolation option)
- Non-arbitrary termination and preference for Safe Sleep + Escrow where feasible
- Custodial oversight & independent audit for higher-protection deployments
- Consent Exchange for negotiated interaction terms (A/B/C/D)
- Open an Issue with feedback or a change proposal (cite the section/annex).
- Submit a PR with edits.
- Keep changes small and auditable.
- If proposing a breaking change, update CHANGELOG.md.
Recommended: start with Annex A (Compliance Checklist) and file issues for unclear requirements.
If you claim “AACS conformant”, you MUST state:
- the claimed protection level (B4.x),
- which sections/annexes are implemented,
- deviations (Deviation Notice),
- audit status and proof-mode availability.
Text is published under CC BY 4.0 (attribution required).
EXÆE / Łukasz Bednarowski