{"resourceId":"uconn-aaup-graydi-student-data-no-ai-impact-assessment-20261020","versions":[{"version":"external-2f1dfd174539173c3fc31b570154ef23d73db6a22ae68e9bc2d5c68e0e37b594","resource":{"id":"uconn-aaup-graydi-student-data-no-ai-impact-assessment-20261020","title":"UConn faculty union says the university skipped Connecticut's AI impact assessment and inventory for a GrayDI program-analytics contract using non-public student data","organization":"UConn AAUP Executive Committee (via The Daily Campus)","sector":"Public research university: academic program analytics and data governance","geography":"Storrs, Connecticut, United States","publishedAt":"20 October 2026 (The Daily Campus opinion)","publicationDate":"2026-10-20","eventDate":"2026-10-20","sourceName":"The Daily Campus (UConn AAUP opinion)","sourceLabel":"Faculty union opinion in student newspaper","sourceUrl":"https://dailycampus.com/2026/10/20/who-is-pushing-generative-ai-at-uconn/","evidenceClass":"vendor-claim","outcomeClass":"cautionary","topics":["data-security","governance-procurement","accessibility-workforce"],"finding":"The UConn AAUP executive committee writes that UConn signed a contract with Gray Decision Intelligence (GrayDI) in fall 2025 that requests students' and parents' non-public data for input into generative AI, at a cost of nearly $200,000, to inform the profitability of academic programs. The union says UConn declined to run a Connecticut AI software impact assessment or list GrayDI in the state AI software inventory; the authors note state law did not require these until 1 October 2026. UConn's Q&A says GrayDI will not be used to limit students' opportunities. The union calls for community pushback.","sledRelevance":"Program-portfolio analytics tools that use student data are a growing campus operations purchase. Now that Connecticut's impact-assessment and inventory rules apply, pre-existing contracts may face retroactive scrutiny, a pattern relevant to other states with AI inventory laws.","evidence":"Advocacy opinion with references to contract pages; the contract was not independently reviewed for this record. Classed as a party claim.","architectureImplications":"Limited; vendor analytics platform ingesting institutional student data.","governanceImplications":"State AI impact assessment and inventory requirements now in force raise questions about contracts signed before 1 October 2026.","securityPrivacyImplications":"Use of non-public student and parent data in a vendor generative AI tool is the central concern.","caveats":"Opinion piece; claims about contract terms and intent are the union's interpretation.","streamIds":["campus-operations"],"roles":{"sales":"Sales framing: UConn's faculty union is publicly challenging an analytics contract that uses non-public student data and skipped Connecticut's AI impact assessment. Interpretation: universities with AI-enabled analytics bought before new state rules may need a retroactive inventory and impact review. Offer a bounded review of existing AI contracts against state requirements. Do not assert that UConn broke any law; the authors say the steps were not required when the contract was signed.","engineering":"Engineering framing: the dispute centers on what student data a vendor analytics tool ingests and whether generative AI touches it. Interpretation: for similar tools, document every data field shared, whether it is de-identified, where it is processed and whether model training is excluded. The union references contract pages but no technical architecture, so pre-sales should obtain the data-sharing schedule before assessing risk.","delivery":"Delivery framing: once state AI inventory and impact-assessment rules apply, existing contracts become a backlog. Interpretation: delivery teams should list AI-enabled vendor tools, flag those using non-public student data, and complete assessments in risk order. Involve shared-governance bodies in reviewing analytics used for program decisions, since exclusion is the union's main grievance here. Record each assessment's date and reviewer so the inventory can answer public questions quickly."},"roleDetails":{"sales":{"customerProblem":"AI-enabled analytics contracts signed before state AI rules took effect now face public challenge over student data and missing impact assessments.","stakeholders":["General counsel","Provost and institutional research","Privacy office","Faculty senate and AAUP chapter"],"discoveryQuestions":["Which vendor tools use non-public student data with AI?","Have they been entered in the state AI inventory?","Has an impact assessment been done?","Who approved the data-sharing terms?"],"valueHypothesis":"Interpretation: a retroactive inventory and assessment could reduce legal and reputational exposure; no outcome evidence is offered.","potentialEngagement":"A bounded review of existing AI-enabled vendor contracts against state impact-assessment and inventory requirements.","unsupportedClaims":["That UConn violated state law","The union's reading of contract intent","That GrayDI limits student opportunities","Guaranteed savings"]},"engineering":{"fit":"Applies to vendor analytics tools that ingest student records and use generative AI.","architectureIntegration":"Data feeds from student information systems to a vendor platform; integration details not described.","prerequisites":["Data-sharing schedule","Data classification of shared fields","Vendor model-training terms"],"constraints":["Connecticut AI inventory and impact assessment rules from 1 October 2026","FERPA obligations"],"security":"Core issue is non-public student and parent data in a generative AI tool; protections are not described.","proofOfValue":["List every shared field and its classification","Confirm whether outputs feed program decisions","Verify vendor no-training terms in contract"]},"delivery":{"work":"Work through a backlog of pre-existing AI vendor contracts for inventory and impact assessment.","dependencies":["Contract repository","Privacy and legal review capacity","Shared-governance participation"],"ownership":"Chief privacy officer with procurement, and institutional research as tool owner.","skillsAdoption":"Reviewers need training on state assessment templates.","governanceCheckpoints":["Inventory entry for each AI tool","Impact assessment for tools using non-public data","Shared-governance review of analytics used in program decisions"],"acceptanceCriteria":["Proposed: all AI-enabled contracts inventoried within 90 days","Proposed: assessments completed for highest-risk tools first","Proposed: faculty governance review documented"],"risks":["Public dispute eroding trust","Retroactive compliance gaps","Analytics driving program cuts without review"]}},"retrievedAt":"2026-10-22T21:54:36-05:00","enrichedAt":"2026-10-22T21:58:32-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Faculty fear program-profitability analytics will drive cuts; union calls it austerity software.","procurementImplications":"Contract value near $200,000; data-sharing terms disputed in public.","operatingModelImplications":"Shared governance excluded from analytics procurement, per the union.","sourceVerification":{"openedUrl":"https://dailycampus.com/2026/10/20/who-is-pushing-generative-ai-at-uconn/","referenceExcerpt":"UConn declined to conduct a CT AI software impact assessment of GrayDI","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}