CSOAI / EU AI Act GPAI Code of Practice · NIST AI RMF · ISO 42001 Annex A
AGI / Frontier-AI Safety — pre-deployment safety cases
AGI-class deployments need a safety case before they ship — not after. CSOAI ships pre-deployment evaluation MCPs that emit signed safety-case evidence packs aligned with Anthropic RSP, OpenAI Preparedness, GPAI Code of Practice, and the UK AISI eval framework.
- Eval-harness MCP — runs cyber/CBRN/persuasion evals + emits signed score sheet
- Capability-elicitation MCP — red-team prompt sets + signed completion logs
- Safety-case generator MCP — auto-drafts the dossier your AISI / National Authority will read
- Care-membrane MCP — runtime alignment guardrails with signed evidence stream
pip install meok-care-membrane-mcp
Covered: GPAI Code of Practice Section II.4 (systemic risk) · UK AISI eval framework · Anthropic RSP · OpenAI Preparedness · ISO/IEC 42005