Research integrity in the age of open access and AI: The view from France
Open access increases the reach and impact of research. That also raises the stakes for trust. Add generative AI to the mix, and questions about authorship, reproducibility, and ethics move to the front of the queue.
A recent roundtable with perspectives from an academic institution, a library, and a publisher made one thing clear: integrity won't hold by default. It needs clear guidelines, consistent training, and active collaboration across the ecosystem.
What integrity means now
France defines research integrity as the rules and values that guarantee honesty and rigour in research. That definition is stable; its application is not. Disciplines interpret and apply it differently, which is why coordination across researchers, institutions, publishers, and libraries is essential.
The French Office for Scientific Integrity (OFIS) highlights this shared responsibility. Policies must be clear, enforceable, and consistently applied, with room for field-specific practice.
Using AI with accountability
AI can accelerate literature reviews, checks, and manuscript prep. It can also introduce bias, blur authorship, and mask data provenance. The solution is not to reject AI, but to set guardrails and keep humans in the loop.
Publishers are integrating AI for quality checks and metadata tasks under strict governance. Libraries are pressing for transparent algorithms and documented data sources so researchers can trust the tools they depend on.
Peer review that earns trust
Open peer review and post-publication commentary increase visibility into the review process. They won't fix every flaw, but they add scrutiny where it matters.
Editors need vigilance to spot anomalies-AI-written reviews, suspicious speed, or coordinated behaviour. Libraries can help researchers understand evolving evaluation practices and what "transparent review" actually means in practice.
Train for culture change
Doctoral training on ethics and integrity is improving, but it often stops there. Senior researchers and supervisors also need ongoing education. Culture shifts when the whole hierarchy is involved.
Libraries run training on citation ethics, plagiarism prevention, and data management, yet these efforts are not always recognized in institutional frameworks. Publishers, operating across jurisdictions, push for harmonized guidance and accessible resources that work across borders.
Corrections without stigma
Retractions should be seen as part of scientific self-correction, not as a scarlet letter. Clear notices and consistent processes help the community understand what happened and why action was taken.
Best practice is to investigate thoroughly, correct the record, and explain the decision. For reference, see COPE's guidance on retractions (Committee on Publication Ethics).
What to do next
- Researchers: State how AI was used, keep full data and code with persistent identifiers, and follow citation norms that make verification easy.
- Libraries: Lead training on data governance, responsible AI use, and research evaluation literacy. Champion adoption of trusted tools with transparent sources.
- Institutions: Extend integrity training to supervisors and senior staff. Recognize and resource library-led programs.
- Publishers: Disclose AI-assisted steps in workflows, strengthen screening, and keep correction policies clear and public.
Practical resources
For ongoing guidance on responsible AI in scientific workflows, explore AI for Science & Research. For national context and frameworks, review the remit and materials from France's OFIS.
Trust scales when every stakeholder does their part-policies that are clear, training that's continuous, and tools that are transparent. Do that consistently, and integrity keeps pace with open access and AI.
Your membership also unlocks:
