Trust Handoff<\/a> posts already enforce explicit permission gating, session capability tracking, and output provenance. Colorado AI Act compliance requires making that instrumentation durable, queryable, and human-readable.<\/p>\n\n\n\nimport os\nimport json\nfrom datetime import datetime\nfrom dataclasses import dataclass, asdict\nfrom typing import Optional\n\n\n@dataclass\nclass ConsequentialDecisionRecord:\n \"\"\"\n Immutable audit record for any agent action that materially\n influences a consequential decision in a Colorado AI Act-\n or EU AI Act-covered domain.\n\n Required fields map directly to the Colorado AI Act's\n transparency and post-modification review obligations.\n \"\"\"\n record_id: str\n timestamp: str\n agent_id: str\n agent_version: str # tracks post-modification review trigger\n decision_domain: str # employment, credit, housing, etc.\n input_summary: str # factors the agent evaluated\n output_summary: str # what the agent recommended or decided\n confidence_score: Optional[float]\n human_review_available: bool # right-to-appeal flag\n human_notified: bool # transparency obligation met flag\n session_capability_flags: list[str] # from the Lethal Trifecta pattern\n\n\nAUDIT_LEDGER: list[dict] = [] # persist to durable storage in production\n\n\ndef record_consequential_decision(\n agent_id: str,\n agent_version: str,\n decision_domain: str,\n input_summary: str,\n output_summary: str,\n session_capability_flags: list[str],\n confidence_score: Optional[float] = None,\n) -> ConsequentialDecisionRecord:\n \"\"\"\n Creates an immutable audit record before any consequential action\n is executed. The record persists regardless of whether the action\n succeeds, fails, or is subsequently appealed \u2014 which is the\n immutability requirement from both Colorado AI Act and EU AI Act.\n \"\"\"\n record = ConsequentialDecisionRecord(\n record_id=f\"cdr_{os.urandom(4).hex()}\",\n timestamp=datetime.utcnow().isoformat(),\n agent_id=agent_id,\n agent_version=agent_version,\n decision_domain=decision_domain,\n input_summary=input_summary,\n output_summary=output_summary,\n confidence_score=confidence_score,\n human_review_available=True, # always True for in-scope deployments\n human_notified=False, # set True after notification is sent\n session_capability_flags=session_capability_flags\n )\n\n AUDIT_LEDGER.append(asdict(record))\n print(f\"[AUDIT LOGGED] {record.record_id} | domain: {decision_domain} | \"\n f\"agent: {agent_id} v{agent_version}\")\n return record\n\n\ndef generate_transparency_notice(record: ConsequentialDecisionRecord) -> str:\n \"\"\"\n Produces the plain-language transparency notice the Colorado AI Act\n requires be provided to any person subject to a consequential decision\n that a high-risk AI system substantially influenced.\n \"\"\"\n return (\n f\"Notice: An AI system (ID: {record.agent_id}, \"\n f\"version {record.agent_version}) contributed to a decision \"\n f\"in the domain of {record.decision_domain} at \"\n f\"{record.timestamp}. \"\n f\"Factors evaluated: {record.input_summary}. \"\n f\"You have the right to request human review of this decision. \"\n f\"Reference: {record.record_id}.\"\n )\n\n\nif __name__ == \"__main__\":\n # Example: an automated loan pre-qualification agent\n record = record_consequential_decision(\n agent_id=\"loan_prequalification_agent\",\n agent_version=\"2.1.4\",\n decision_domain=\"financial_services_credit\",\n input_summary=\"credit score 712, DTI ratio 0.31, employment 3yr stable\",\n output_summary=\"pre-qualified for $280,000 at 6.8% APR \u2014 recommended proceed\",\n session_capability_flags=[\"PRIVATE_DATA_ACCESS\", \"AUTHENTICATED_SYSTEM\"],\n confidence_score=0.87\n )\n\n notice = generate_transparency_notice(record)\n print(f\"\\n[TRANSPARENCY NOTICE]\\n{notice}\")<\/code><\/pre>\n\n\n\nThe agent_version<\/code> field is doing specific compliance work: it creates an automatic post-modification review trigger in your audit ledger. Whenever that version string changes, any consequential decision made before and after the modification is traceable \u2014 which is exactly what the Colorado AI Act’s post-modification review obligation requires.<\/p>\n\n\n\n
\n\n\n\n
![\"Colorado](\"https:\/\/www.theagenticprotocol.com\/wp-content\/uploads\/2026\/07\/Sleep_stage_architecture_chart_202607031613-1024x572.jpeg\")
<\/figure>\n\n\n\n