{"resourceId":"ec-era-living-guidelines-v4-agentic-ai-peer-review-20261015","versions":[{"version":"external-0428d74b4c5bc87a05a1c468436ccf5bb195c0174e48feaf43cfabd61f5bde10","resource":{"id":"ec-era-living-guidelines-v4-agentic-ai-peer-review-20261015","title":"European Commission updates ERA Living Guidelines on generative AI in research: agents stay tools, merit judgements stay human, institutions should offer governed AI","organization":"European Commission, Directorate-General for Research and Innovation (ERA Forum)","sector":"Research funders and research-performing organisations","geography":"European Union / European Research Area (non-binding guidance; transferable to U.S. research offices only as a reference model)","publishedAt":"15 October 2026 (fourth version; Commission news item 16 October 2026)","publicationDate":"2026-10-15","eventDate":null,"sourceName":"European Commission","sourceLabel":"ERA Living Guidelines on the responsible use of generative AI in research, fourth version (PDF)","sourceUrl":"https://research-and-innovation.ec.europa.eu/document/download/2b6cf7e5-36ac-41cb-aab5-0d32050143dc_en?filename=ec_rtd_ai-guidelines.pdf","evidenceClass":"standards-guidance","outcomeClass":"emerging","topics":["governance-procurement","data-security"],"finding":"The fourth version of the ERA Living Guidelines (published 15 October 2026; first version March 2024, previous revision May 2026) adds agentic AI and expands peer-review and proposal-evaluation guidance. It says AI agents must be used as research tools, not independent scientists, with human validation of agent-generated hypotheses, data processing and conclusions. Researchers should disclose substantial AI use with oversight and verification methods, and consider retaining prompts, logs, model versions and parameters because outputs are non-deterministic. Reviewers and evaluators must not delegate scientific-merit judgement to AI or put unpublished proposals into non-secured tools. Research organisations should, where possible, provide locally hosted or cloud tools they govern themselves.","sledRelevance":"U.S. research universities face the same questions about agents, disclosure and AI in internal review, and many collaborate with EU partners or Horizon-funded projects. The document is a ready reference for a research-office policy update, though it is non-binding and EU-specific.","evidence":"Opened the fourth-version PDF and the 16 October Commission news item. Read the recommendations for researchers, research organisations and funders, the agentic AI and peer-review sections, and the revision history (fourth version published 15 October 2026; next revision foreseen Q3 2027).","architectureImplications":"Institution-governed AI services (local or cloud with clear data management) are recommended for sensitive research data; agent runs should keep prompts, logs, model versions and parameters where needed to reproduce or audit work.","governanceImplications":"Puts accountability on the researcher for every agent step, sets a disclosure threshold for substantial use, and bars AI from drafting evaluation reports or judging merit in peer review and funding evaluation.","securityPrivacyImplications":"Unpublished manuscripts, proposals and review reports should not go into non-secured AI tools; users should check whether inputs may be reused for training. Uploading personal data still needs a lawful basis.","caveats":"Principles, not an implementation standard or law; no outcome data that the recommendations reduce misconduct or bias; EU legal context differs from U.S. federal and state rules; interpretation of what counts as substantial use is left to disciplines and institutions.","streamIds":["research"],"roles":{"sales":"Lead with policy readiness, not tools. The Commission now says agents are research tools, merit judgements stay human, and institutions should offer governed AI for sensitive data. Research VPs and integrity officers updating local policy can use this as a reference. A bounded engagement compares current campus policy and tools against these recommendations. Interpretation: this is non-binding EU guidance, so it supports a gap review, not claims of compliance or reduced misconduct.","engineering":"Treat the guidance as requirements for a governed research-AI service. It points to institution-run or contracted tools with clear data handling, run records (prompts, logs, model versions, parameters) for agent work, and a block on unpublished proposals entering public tools. Pre-sales should map existing campus AI services to these points, including where data is processed and whether inputs train models. Interpretation; the document gives no technical standard or reference architecture.","delivery":"Frame delivery as turning principles into campus procedure. Work would cover a disclosure template, a rule for retaining agent run records, and a peer-review and internal-competition policy that keeps merit judgement human. The research office and integrity office own it with IT. Interpretation: success should be measured by policy adoption and audit findings, since the guidance itself reports no outcome evidence."},"roleDetails":{"sales":{"customerProblem":"Research offices lack clear local rules for AI agents, AI disclosure and AI use in internal review and grant competitions.","stakeholders":["Vice president or provost for research","Research integrity officer","Graduate school dean","Research IT and information security"],"discoveryQuestions":["Does campus policy say who is accountable when an AI agent performs research steps?","Are internal grant reviewers told what AI may and may not do?","Which AI tools are approved for unpublished or sensitive research data?"],"valueHypothesis":"A policy gap review against this guidance may shorten the time to a defensible campus position on agents and AI in review.","potentialEngagement":"A four- to six-week policy and tooling gap assessment against the fourth-version recommendations, ending in draft disclosure and review rules.","unsupportedClaims":["Guaranteed reduction in research misconduct or bias","Legal compliance in U.S. jurisdictions","Evidence that any specific tool meets the guidance"]},"engineering":{"fit":"Fits institutions designing or reviewing a governed research-AI service; limited to principles rather than technical controls.","architectureIntegration":"Map campus AI services to data-handling, retention and run-record expectations, and add logging of model versions and parameters for agent workflows.","prerequisites":["Inventory of AI tools in research use","Data classification for unpublished and personal research data","Existing logging capability for AI services"],"constraints":["Non-binding EU guidance; U.S. federal, state and sponsor rules govern locally","No technical standard or certification is defined","Disciplines and journals may set different disclosure thresholds"],"security":"Keep unpublished manuscripts, proposals and review reports out of non-secured tools and confirm whether vendors retain or train on inputs.","proofOfValue":["Trace one agent-assisted analysis end to end with prompts, logs and model versions retained","Confirm a governed tool blocks training reuse of inputs in contract and configuration","Test that internal-review guidance is applied in one grant competition cycle"]},"delivery":{"work":"Draft disclosure templates, agent run-record retention rules and AI-in-review rules for internal competitions, and align approved tools with them.","dependencies":["Research integrity office sign-off","IT and security review of approved tools","Faculty consultation"],"ownership":"Research office with the research integrity office, supported by research IT.","skillsAdoption":"Training for researchers, reviewers and research administrators on disclosure and on what AI may do in review.","governanceCheckpoints":["Policy approval by research leadership","Annual review aligned with guideline revisions","Audit of a sample of disclosures"],"acceptanceCriteria":["Proposal: disclosure template adopted by the research office within one term","Proposal: every approved research-AI tool documented for data handling and training reuse","Proposal: internal-competition reviewers attest to AI-use rules each cycle"],"risks":["Policy that is too broad to apply consistently","Researchers using unapproved public tools anyway","Divergence from sponsor or journal rules"]}},"retrievedAt":"2026-10-18T21:59:55-05:00","enrichedAt":"2026-10-18T21:59:55-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Asks research organisations to provide training on responsible AI use and to track use; institutions will need staff time for training and policy upkeep.","procurementImplications":"Favours bespoke or institution-governed AI services over public instances; procurement should ask where systems run, who operates them and whether prompts are retained or reused.","operatingModelImplications":"Research offices, integrity offices and IT share ownership: policy, disclosure templates, governed tools and review rules for internal grant competitions.","sourceVerification":{"openedUrl":"https://research-and-innovation.ec.europa.eu/document/download/2b6cf7e5-36ac-41cb-aab5-0d32050143dc_en?filename=ec_rtd_ai-guidelines.pdf","referenceExcerpt":"AI agents must be deployed strictly as advanced research tools, not independent scientists.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}