Guidelines and procedures for the responsible use of artificial intelligence across all SAM activities.
Why these guidelines exist
The Guidelines on the use of AI in the SAM arise from acknowledging the opportunities offered by artificial intelligence, whilst accounting for the risks and challenges facing its use for the purposes of scientific advice for policymaking.
Through this guidance, the Scientific Advice Mechanism remains open to innovation and greater efficiency, while safeguarding scientific excellence and public trust through clear standards of transparency, accountability, and human oversight. It also takes into account the specific structure of the SAM and the variety of stakeholders involved in its processes.
These guidelines apply to everyone involved in SAM tasks and processes: the Group of Chief Scientific Advisors, SAPEA, the SAM Secretariat, and all experts participating in SAM activities, including Working Group members, workshop participants, and peer reviewers.
The guidelines map the full range of tasks carried out within the SAM (scientific, communication-related, and administrative), from output drafting, literature searching and screening, peer review and policy mapping, to email drafting, media monitoring, event organisation, translation, and expert identification. This mapping (Appendix 4 of the guidelines) grounds the principles in the day-to-day reality of the Mechanism’s work.
The general principles
- Personal Responsibility and Accountability: Users are responsible and accountable for understanding both the capabilities and risks of using AI tools and ensuring the integrity and accuracy of the content produced.
- Human-Centred use of AI: AI should be used to complement, rather than replace, human judgment and expertise.
- Review and Human Control: Any content delivered with the support of AI tools should be critically reviewed before use.
- Transparency & Reporting: The usage of AI tools, including the tasks in which they are applied, shall be recorded in the interest of transparency.
- Data Handling & GDPR: AI tool users shall reflect on the type of data entered into an AI tool – please refer to Appendix 2 Data Handling.
- Intellectual Property Compliance: AI tool users are responsible for ensuring, to the best of their ability, that that copyright rules are respected when using content delivered with the support of AI.
- Environmental Footprint: Individuals shall be mindful of their environmental footprint when using AI tools.
How AI tools are assessed
Any of the three components of the SAM (the GCSA, SAPEA, or the SAM Secretariat) can request that a specific AI tool be assessed by the SAPEA Task Force on the use of AI in the SAM.
The assessment is qualitative and carried out from the perspective of a typical, non-expert end user. It checks the tool against the general principles, weighs opportunities and risks, considers costs and usage plans, and compares the tool’s efficiency and effectiveness with the equivalent manual task. The resulting internal report is shared with all three components of the SAM.
Two things remain constant:
- The final decision rests with the individual. The assessment supports reflection; it does not replace personal responsibility.
- Restrictions are proportionate. Each SAM component may independently restrict a tool within its own remit, but a full ban requires a joint decision by all three components.
Assessments are reviewed periodically to keep pace with the fast-moving AI landscape.
What data can go into an AI tool?
Not all data is equal. The guidelines ask every user to pause before entering information into an AI tool, and to treat the following with particular caution:
- Sensitive personal data (names, addresses, contact details) and sensitive data pertaining to communities or territories
- Unpublished copyrighted or IP-protected works
- Applications, reviews, and their content
- Key internal documents, invoices, forms, and business secrets
See Appendix 2 of the guidelines for full details.
How the guidelines were developed
EASAC steered the process leading to the publication of these guidelines, which included evidence collection, content development, expert engagement, and exchanges with a wider community of practice.
The process combined:
- A Working Group of six experts convened by SAPEA to draft the guidelines
- A Task Force of SAPEA staff that mapped SAM tasks and identified tools for assessment
- A literature review by Cardiff University
- Three workshops held under the Chatham House Rule with academics, policy officers, and other stakeholders, covering scientific processes, administrative and operational processes, and communication activities
The final version was approved by all three components of the Scientific Advice Mechanism.
A living document
A review of the guidelines is foreseen in 2027, to incorporate new developments in AI, account for evolving regulatory standards, and reflect the SAM’s continuous reflection on these matters.
Related resources
- Guidelines and procedures for the use of AI in the SAM (full PDF)
- SAPEA (2026): The Potential of AI in Policymaking, Scientific Advice and Evidence Review: Literature Review
- SAPEA (2024): Successful and timely uptake of artificial intelligence in science in the EU
- ALLEA European Code of Conduct for Research Integrity (2023)
- The EU Artificial Intelligence Act (Regulation (EU) 2024/1689)