Council City: Municipal AI Procurement and Boundary Governance
Council City is the municipal application of the Council of AI's monitored containment framework. Local authorities deploying AI in public-facing services — benefits administration, planning decisions, social-care triage — face a governance gap: they lack the in-house AI-safety expertise that national regulators assume, yet their procurement decisions carry direct consequences for vulnerable populations. Council City addresses this by providing a turnkey measurement board for municipal AI procurement: a 14-axis benchmark suite that runs on the authority's own infrastructure (or a governed cloud instance), producing signed measurement credentials without requiring the authority to staff an AI-safety lab.
The model is 'measure, don't trust': every AI system proposed for deployment is benchmarked against the same axes the Council of AI uses for frontier-lab governance — including sandbox-escape detection (Art-5 axis), care-obligation adherence (Care axis), and third-party risk (MCP/OSS axes). The measurement credential is signed on the authority's behalf and published alongside the procurement decision. This shifts the municipal conversation from 'Is this AI safe?' (unanswerable without measurement) to 'Here is the independent measurement record — does it meet our threshold?' (answerable, auditable, and defensible in judicial review).
Council City is designed to run as a governed pilot programme for UK and EU local authorities, using the same measurement board and public-service procurement mapping described above.
Standards and sources referenced
- Council of AI Monitored Containment Framework
- Council City Pilot Programme — DRCF Sandbox Cohort 2
- UK Local Government AI Procurement Guidelines (DLUHC, 2026)
- EU Digital Europe Programme — AI Governance in Public Services (DIGITAL-2026-AI-GOV)