{"resourceId":"detroit-citywide-ai-policy-copilot-chat-only-20261005","versions":[{"version":"external-0a65ff8824356db16561546e8dea568a4a6ce895f0f37b4559ef34d0e301e006","resource":{"id":"detroit-citywide-ai-policy-copilot-chat-only-20261005","title":"Detroit's citywide AI policy allows only Microsoft CoPilot Chat, bans AI in employee evaluations, and requires disclosure and fact-checking","organization":"City of Detroit (Department of Innovation and Technology)","sector":"Municipal AI acceptable-use policy","geography":"Detroit, Michigan, United States","publishedAt":"October 5, 2026 (BridgeDetroit); summary republished by Detroit Newsroom October 7, 2026","publicationDate":"2026-10-05","eventDate":null,"sourceName":"BridgeDetroit","sourceLabel":"Nonprofit local news explainer","sourceUrl":"https://www.bridgedetroit.com/how-is-detroit-using-ai-we-dug-into-the-citys-policy/","evidenceClass":"independent-reporting","outcomeClass":"emerging","topics":["governance-procurement","data-security","knowledge-work","accessibility-workforce"],"finding":"Detroit's AI policy, in effect since this summer for all departments, vendors, interns, consultants and volunteers, permits only Microsoft CoPilot Chat as a text LLM. It bars AI-made images of real events, AI in evaluations of people or programs such as employee reviews, and AI handling sensitive or identifying data without human review; staff must disclose AI use and fact-check outputs. Violations are handled through the employee handbook up to discharge, and new tools go through DoIT review. The CIO's main stated worry is AI literacy.","sledRelevance":"A large U.S. city chose a single-approved-model policy tied to its surveillance-oversight ordinance, a model other cities with similar ordinances can compare against.","evidence":"Explainer based on the policy and a CIO interview; no compliance or usage data.","architectureImplications":"A single approved LLM inside the Microsoft tenant means blocking or discouraging other generative AI services and routing new-tool requests through DoIT.","governanceImplications":"Links AI to the Community Input Over Government Surveillance ordinance and employee discipline; drafting involved Law, HR, Procurement, police and fire, and the Council's Legislative Policy Division.","securityPrivacyImplications":"Prohibits sharing data unsafe for public disclosure and AI processing of personal data without human review; 'Detroit's data is Detroit's data.'","caveats":"Published Oct. 5, before the last run, but not in the archive and amplified Oct. 7; the policy text itself was not opened separately; no audit of compliance.","streamIds":["local-government"],"roles":{"sales":"Sales framing: Detroit has standardised on one approved LLM and binds vendors to its policy, so any AI-enabled product must fit through DoIT review and the surveillance-oversight ordinance. Useful discovery covers which upgrades to existing systems quietly add AI and how disclosure is handled. The CIO's stated concern is literacy, which points toward training support rather than new tools. Avoid implying the city is shopping for alternative models; the policy deliberately narrows the field, and no usage figures are public.","engineering":"Engineering framing: The technical consequence is a Microsoft-centred AI footprint with an intake process for anything else. Engineers should expect questions about whether a product's AI features can be disabled, where data is processed, and whether outputs can be labelled for disclosure. Constraints include the ban on AI processing personal data without human review and the surveillance ordinance. A proof of value could test whether an AI-enabled system upgrade can be configured to keep sensitive data out of model calls.","delivery":"Delivery framing: Delivery here is compliance and literacy work. DoIT owns review of new tools, so projects need a request package showing data flows, human review steps and disclosure wording. Interpretation: an inventory of AI features already embedded in vendor upgrades would help the city apply its own policy. Training should cover fact-checking and disclosure in daily work, given the CIO's concern. Measure policy awareness and request turnaround rather than claiming productivity gains the source does not report."},"roleDetails":{"sales":{"customerProblem":"The city must keep AI use within one approved tool while vendors add AI features to existing systems.","stakeholders":["CIO, Department of Innovation and Technology","Law Department","Human Resources","Procurement","City Council Legislative Policy Division"],"discoveryQuestions":["Which existing vendor upgrades include AI features?","How long does DoIT review of a new AI tool take?","How is AI use disclosed in documents?","What AI literacy training exists for staff?"],"valueHypothesis":"Clearer intake and training could reduce policy violations; no baseline is reported.","potentialEngagement":"An inventory of AI features in current vendor systems mapped to the policy's prohibitions.","unsupportedClaims":["That other LLMs will be approved soon","Measured productivity gains","That compliance is currently high or low"]},"engineering":{"fit":"Fits Microsoft-tenant deployments and products whose AI features can be configured or disabled.","architectureIntegration":"CoPilot Chat inside the city tenant; DoIT intake for other tools; disclosure labelling and human review for personal data.","prerequisites":["Microsoft tenant controls","Data classification","DoIT review workflow","Vendor documentation of AI features"],"constraints":["Single approved LLM","Surveillance ordinance compliance","No AI in personnel evaluations","No AI handling personal data without human review"],"security":"Keep data unsafe for public disclosure out of AI prompts and require human review where personal data is involved.","proofOfValue":["Configure one vendor upgrade with AI features disabled or scoped, and verify","Test disclosure labelling in a document workflow","Check logs for use of unapproved AI services"]},"delivery":{"work":"Operate the policy: tool intake, vendor AI inventory, disclosure practice and staff literacy.","dependencies":["DoIT review capacity","Vendor cooperation","HR for enforcement","Training resources"],"ownership":"Department of Innovation and Technology (CIO).","skillsAdoption":"Literacy training on fact-checking, disclosure and data limits for employees, vendors and volunteers.","governanceCheckpoints":["New-tool review decision","Annual policy review","Council public comment"],"acceptanceCriteria":["All vendor systems with AI features inventoried","Median review time for AI requests published","Share of staff completing AI training reported"],"risks":["Shadow AI use outside CoPilot Chat","Unreviewed AI features in vendor upgrades","Literacy gaps among staff and residents"]}},"retrievedAt":"2026-10-07T22:12:17-05:00","enrichedAt":"2026-10-07T22:21:14.164-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Bans AI in employee reviews; the CIO flags uneven AI literacy among staff and residents.","procurementImplications":"Vendors and consultants are bound by the policy, so contract terms must reflect the disclosure and data limits.","operatingModelImplications":"DoIT staff, including the CIO, review each request for new AI tools or AI-enabled upgrades.","sourceVerification":{"openedUrl":"https://www.bridgedetroit.com/how-is-detroit-using-ai-we-dug-into-the-citys-policy/","referenceExcerpt":"Currently, the only text-based model like this (otherwise known as a Large Language Model or LLM) city employees can use is Microsoft CoPilot Chat","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}