{"resourceId":"gatech-ai-academic-research-policy-202606","versions":[{"version":"external-141b06f08e475be9d0aba4a671d4459560a81fbb03a2ed0b04461b056e88de6e","resource":{"id":"gatech-ai-academic-research-policy-202606","title":"Georgia Tech interim AI Academic & Research policy: approved tools, protected-data limits, human oversight, disclosure","organization":"Georgia Institute of Technology (Office of the Provost / EVPR)","sector":"Higher education research administration","geography":"United States (Atlanta, GA)","publishedAt":"June 2026 (effective); review December 2026","publicationDate":"2026-06-01","eventDate":null,"sourceName":"Georgia Tech Policy Library","sourceLabel":"University policy (Georgia Tech)","sourceUrl":"https://policylibrary.gatech.edu/artificial-intelligence-ai-academic-research-contexts","evidenceClass":"standards-guidance","outcomeClass":"emerging","topics":["governance-procurement","data-security","operating-model"],"finding":"Georgia Tech’s interim policy governing AI in Academic and Research Contexts (effective June 2026) requires community members to use Institutionally Approved or Local-POC Limited Use AI tools for academic/research work; prohibits submitting Protected Data or PII into tools outside approved scope; mandates human oversight/verification of AI outputs before reliance; and requires appropriate disclosure of AI use in scholarly communications. An AI Governance Officer and Committee coordinate tool approval and points of contact. Why now: as campuses expand AI/HPC capacity and research agents show low claim-reproduction rates, research-admin control layers become operationally binding—not optional guidance.","sledRelevance":"Research administration adopting AI agents and shared AI tools needs enforceable approval, data-handling, oversight, and disclosure controls that transfer to other R1 governance designs.","evidence":"Georgia Tech Policy Library page opened via HTTPS; effective date June 2026, tool-approval, protected-data, human oversight, and disclosure provisions taken from the opened policy text.","architectureImplications":"Institute and local AI registries; Local AI Points of Contact; separation of academic/research vs administrative AI policies; training requirements via AI Governance Program.","governanceImplications":"Central AI Governance Committee plus unit POCs; annual review; enforcement via existing academic and research misconduct channels; funder/publisher policies still apply.","securityPrivacyImplications":"Hard stop on Protected Data/PII in unapproved tools; prompt-injection awareness; monitoring/audit trails referenced to companion administrative AI policy.","caveats":"Interim policy without published compliance metrics; effective June 2026 (older than last edition) included for newly relevant control-layer why-now; does not evaluate scientific agent performance.","streamIds":["research"],"roles":{"sales":"Sales framing: Research-admin and compliance buyers need a concrete university control model for AI tools—approved registries, protected-data bans, human oversight, and disclosure—not another model license. Georgia Tech’s interim Academic & Research AI policy is a transferable pattern for discovery. Do not claim measured risk reduction; this is policy design evidence, newly salient as agent workflows and campus AI capacity expand.","engineering":"Engineering framing: Fit requires wiring tool registries, identity-aware allowlists, and data-classification checks into research platforms and agent gateways so Protected Data cannot reach unapproved tools. Prerequisites include data classification, Local AI Points of Contact, and documented approval workflow. This interpretation is bounded to the retrieved source evidence and should not be generalized beyond stated limits. This interpretation is bounded to the retrieved source evidence and should not be generalized beyond stated limits.","delivery":"Delivery framing: Stand up an AI Governance Officer and Committee, publish the approved-tool list, train PIs on disclosure and verification, and route unapproved tools through Limited Use review. Acceptance includes live registries, assigned training, and documented protected-data path. Risks include shadow AI tools and disclosure theater. This interpretation is bounded to the retrieved source evidence and should not be generalized beyond stated limits. This interpretation is bounded to the retrieved source evidence and should not be generalized beyond stated limits."},"roleDetails":{"sales":{"customerProblem":"Research units adopt AI tools and agents faster than institutional approval, data, and disclosure controls can keep up.","stakeholders":["Provost / EVPR","Research integrity / ORIA","CIO / OIT governance","College AI points of contact","PIs and lab managers"],"discoveryQuestions":["Is there an institute AI registry and Limited Use path?","How is Protected Data defined and enforced in AI tools?","Who verifies AI outputs before scholarly use?","What disclosure standard applies to grants and manuscripts?"],"valueHypothesis":"An interim academic/research AI policy with approved tools, data limits, human oversight, and disclosure can reduce uncontrolled AI use while enabling sanctioned tools.","potentialEngagement":"Gap assessment of campus AI tool inventory against an approved-registry and protected-data control model.","unsupportedClaims":["Measured reduction in research misconduct","Complete prevention of shadow AI","Scientific quality improvement from policy alone","Identical fit to every state system"]},"engineering":{"fit":"Policy-backed allowlist of AI tools with local Limited Use exceptions and data-class enforcement for research systems.","architectureIntegration":"Connect IdP, data classification labels, AI registries, and research platforms so unapproved tools cannot receive Protected Data.","prerequisites":["Data classification scheme","AI Governance Officer/Committee","Local AI POCs","Tool approval standard","Training modules"],"constraints":["Interim policy","No published compliance KPIs","Companion admin policy dependency","Student personal-tool carve-outs"],"security":"Block Protected Data/PII into non-approved scopes; monitor prompt-injection and audit trails per companion IT policy.","proofOfValue":["Share of research AI use on approved tools","Documented Limited Use reviews","Training completion","Disclosure appearing in sample scholarly outputs"]},"delivery":{"work":"Publish registries, appoint POCs, roll training, embed disclosure guidance in grant/manuscript workflows.","dependencies":["OIT registry platform","ORIA / research integrity","College deans for POCs","CTL for instructional guidance"],"ownership":"AI Governance Program coordinates; Provost/EVPR own academic/research decisions; units own local POCs.","skillsAdoption":"Train PIs and students on verification, disclosure, and approved-tool selection; train POCs on Limited Use triage.","governanceCheckpoints":["Annual policy review","Tool approval audits","Misconduct process linkage"],"acceptanceCriteria":["Approved and Limited Use registries published","Protected-data guidance live","POC network designated"],"risks":["Shadow SaaS AI tools","Uneven college enforcement","Disclosure without verification","Policy lag vs agent capabilities"]}},"retrievedAt":"2026-09-26T21:59:45-05:00","enrichedAt":"2026-09-26T22:08:00-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Policy lists accessibility among guiding principles and requires education offerings; Local POCs are the workforce layer that makes registries usable.","procurementImplications":"Prefer Institutionally Approved AI tools; route new research tools through approval standard before contract—especially those touching Protected Data.","operatingModelImplications":"Central governance with distributed college POCs is the operating model for research AI enablement under control.","sourceVerification":{"openedUrl":"https://policylibrary.gatech.edu/artificial-intelligence-ai-academic-research-contexts","referenceExcerpt":"either an (1) Institutionally Approved AI Tool or a (2) Limited Use AI Tool as designated","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}