The European Commission has just updated its guidelines on how researchers should use generative AI. The update warns institutions about “hidden prompts”, instructions to AI systems hidden from human oversight, and flags the risks of third parties using AI tools, including AI notetakers, in research meetings.
The recent update was published by the Commission’s Directorate-General for Research and Innovation. The Commission itself describes the revision as “limited and technical”, which is honest framing. The bigger rules were laid down in March 2024, when the original Living Guidelines were published with input from European Research Area Forum countries and stakeholders.
The 2026 update does add two specific concerns worth noting. First, “hidden prompts”, where instructions for AI systems are concealed from human oversight. Second, the risks when third parties use AI during research meetings, think AI notetakers joining a call without everyone knowing. Both reflect what happens when AI gets embedded in everyday workflows faster than the rules catch up.
But the substance, the actual red lines, the actual obligations, those come from the 2024 document. So let me walk you through what the EU’s research AI rulebook actually says.
Where researchers cannot use AI
The guidelines tell researchers to refrain from using generative AI tools in sensitive activities such as peer reviews or evaluations.
Peer review is the immune system of academic research. It’s where weak science gets filtered out. Automate that process with AI, and you undermine it. A generative AI tool cannot assess novelty or methodological appropriateness the way a human expert can. And it could introduce bias at scale.
There is also a confidentiality problem. Peer review involves unpublished research shared in confidence. Upload that manuscript to a third-party AI tool, and you have broken that trust immediately. The guidelines also tell researchers to use AI while respecting privacy, confidentiality, and intellectual property rights. In peer review, those three things are non-negotiable.
Who is responsible
The guidelines do not put it all on individual researchers. Research organisations should facilitate the responsible use of generative AI and actively monitor how these tools are developed and used within their organisations.
Actively monitor. That means universities and research institutes need to know which AI tools their researchers are using, for what purposes, and with what data. It means internal approval processes. Risk assessments. Training that goes beyond a one-off lunchtime session. And a clear policy that tells staff what they can use, what they cannot, and where to go when they are not sure. Publishing a PDF and calling it governance does not cut it.
Here is the catch. Commissioner Iliana Ivanova said when the original guidelines launched in 2024, “Generative AI can hugely boost research, but its use demands transparency and responsibility. These guidelines aim to uphold scientific integrity and preserve public trust in science.” Strong words. But the guidelines are non-binding. They are a supporting tool. Compliance is voluntary. The EU AI Act, by contrast, is legally enforceable, and research institutions using AI in ways that affect fundamental rights could fall under its high-risk categories. So you have two parallel systems, one binding, one voluntary, and institutions need to make sure they are working for both, rather than just one.
The detection problem
Now, a reality check from the ground. The European University Institute, one of Europe’s leading research institutions, has stated openly that it does not use anti-plagiarism tools to detect AI-generated content. The reason? Shortcomings in the technology and the risk of false positives. Instead, the EUI relies on individual responsibility.
Read that again. One of Europe’s leading research institutions has effectively said, ” We cannot reliably detect AI use, so we are trusting people to self-report.”
The issue here is the technology. AI detection tools flag human writing as AI-generated and miss actual AI-generated text. In academia, false positives can destroy careers. So institutions are stuck. They cannot police AI use effectively, and they cannot ignore it either.
If you are listening to this and you do not work in academic research, do not switch off. The challenge facing universities is the challenge facing every organisation. How do you maintain quality when AI is embedded in knowledge work? How do you create transparency when the tools themselves are opaque? And how do you enforce standards when detection is unreliable?
The lesson from Europe’s research guidelines is this. Governance without enforcement is an aspiration. If you are building AI policies in your organisation, make sure they connect to something real. A contractual obligation. A disciplinary process. A link to a binding law like the AI Act. Otherwise, you are hoping AI governs itself.
If you want to build your AI literacy and skills, I can help
📚 Grab AI Playbook on Amazon, the perfect podcast companion.
📰 Subscribe to Human in the Loop, my LinkedIn Newsletter
🎓 Take the CPD Accredited AI Playbook Diploma
📞 Book a call at digitaltraining.ie
Sources
- European Commission, Updated ERA living guidelines on the responsible use of generative AI in research, 8 May 2026: research-and-innovation.ec.europa.eu (updated guidelines news)
- European Commission, Guidelines on the responsible use of generative AI in research developed by the European Research Area Forum, 20 March 2024: research-and-innovation.ec.europa.eu (original guidelines news)
- European Commission, ERA Living Guidelines on the Responsible Use of Generative AI in Research (PDF): Download full PDF
- European University Institute, EUI Guidelines for the Responsible Use of Artificial Intelligence for Research, approved 15 May 2024: eui.eu (EUI AI guidelines PDF)
- EU AI Act overview: digital-strategy.ec.europa.eu (regulatory framework for AI)
Key Statistics
- ERA Living Guidelines first published 20 March 2024, updated 8 May 2026.
- Commission describes the 2026 update as “limited and technical”.
- New additions: guidance on “hidden prompts” and risks of third parties using AI in research meetings.
- Original guidelines developed with input from European Research Area Forum countries and stakeholders.
- Guidelines are non-binding. The EU AI Act is a binding law.
- The European University Institute does not use AI detection tools, citing shortcomings and false positives.
- EUI guidelines approved by the Academic Council on 15 May 2024.
About the Podcast Host
AI and Communications Specialist | Author | Podcaster | International Speaker
Joanne Sweeney is a recognised leader in digital marketing and digital transformation, specialising in the practical application of AI to enhance communication strategies. The founder of Digital Training Institute, Joanne, empowers leaders and communications professionals to leverage AI responsibly and effectively, ensuring transparency, trust, and tangible results.
She completed the Harvard Kennedy School Executive Programme in AI and Public Policy this summer before launching her fourth book AI Playbook, in September. Joanne has also published two editions of Public Sector Marketing Pro, which were preceded by her first title, Social Media Under Investigation Law Enforcement and the Social Web.
Her books have become go-to resources for governments, public sector and policing agencies, as well as marketing and communications professionals, all navigating the challenges of the changing Internet.
Joanne is the host of the AI SIX Podcast, a daily six-minute podcast that provides actionable insights, expert interviews, and practical advice to help professionals navigate the new world of work.
