The Artificial Intelligence in Language Education (AI Lang) workshop

On the 25th and 26 November 2025, I organised –together with a dream team of colleagues– a two-day workshop for Artificial Intelligence in Language Education (AI Lang) project at the European Centre for Modern Languages of the Council of Europe. Bringing colleagues from all over Europe together in Graz (my former academic home) was a welcome opportunity to step back from the day-to-day production of resources and reflect, collectively, on what we are trying to build: principled, ethically grounded, and genuinely useful support for the meaningful use of AI in language education.

In what follows, I sketch the aims of the workshop and the broader context of the AI Lang project, before turning to the substantive work we undertook, from refining our draft guidelines to piloting the Moodle-based professional development course, and also strengthening the emerging community around this initiative. I also reflect on the value of events like this for building a community of practice around responsible AI use in language education, and I outline a few ways to stay connected or get involved.

If you’re curious to see how all this fit together, perhaps you’d like to read on.

The AI Lang project

The Artificial Intelligence in Language Education project is a four-year initiative (2024-2027) by the European Centre of Modern Languages of the Council of Europe, which aims to respond to the way AI is reshaping language education.

Together with my colleagues Anne-Laure Dubrac, Liz Polzleitner and Aleksandra Ljalikova, and with the invaluable help of our associate members Jérémie Séror and Merilyn Meristo, as well as our fellow Konstantina Alevizou, we have been collectively reflecting on how language teachers can use AI to support their practice, while making sure that this use is safe, ethical, purposeful, efficient, reflective and critical.

As the project develops, we intend to produce:

• a set of structured guidelines for using AI in language teaching and learning, which will supplement existing guidance for general education;
• a Moodle-based course which language teachers can use to strengthen their AI literacy.

We strongly believe that these resources should not be in the form of top-down guidance. Rather, they should be open-ended and responsive to the needs of actual practicing teachers. One of the ways in which we try to ensure this is by making sure that we continuously work with teachers and experts who can help us interrogate our assumptions, test the practicality of our ideas, and point out blind spots we may not otherwise notice. Workshops such as the one in Graz are central to this effort: they give us the opportunity to share work-in-progress, gather nuanced feedback, and refine our materials in dialogue with the people who will ultimately use them.

The AI Lang workshop

The November 2025 workshop aimed to do exactly this. Over two days, we invited participants to critique the draft guidelines, explore the beta version of the Moodle course, and reflect with us on what responsible AI use might look like in diverse educational contexts.

In the workshop, we were very proud to host 41 participants from all the member-states represented at the ECML, from Armenia to Iceland and from Norway to Spain, as well as representatives from partner organisations such as EAQUALS and CercleS. We were also delighted to be joined by representatives from other ECML projects, such as ARPIDE and ICT-Rev.¹

Such a diverse group of people was a fantastic opportunity to practice what we preach about multilingualism: throughout the workshop we all made a conscious effort to balance the use of both our project languages: English and French. This was not always straightforward: for instance, many of our working documents are only available in English at this moment. But we did create discussion groups in both English and French, and we did try² to speak both languages in the plenary sessions. The ECML’s simultaneous interpretation team played an essential role here, enabling participants to contribute fully in the language that best supported their thinking.

Refining the guidelines for the use of AI in Language Education

The first day of the workshop focused on our structured guidelines for the use of AI in Language education. This was an intense day of work, involving a three-round World Café activity, and alternating rounds of group-work and plenary feedback. During these activities, participants had the opportunity to read our guidelines, comment on them, and use them to evaluate authentic cases of AI use from various settings across Europe and neighbouring countries.

A highlight of this day was when we fed all the notes we took, and the transcripts from the discussions into NotebookLM, and used this to produce a series of five podcasts that summarised the main findings of the discussion. We will, of course, analyse these data properly, but the activity was useful in demonstrating how to use AI responsibly by eliciting consent, protecting anonymity, and prompting effectively. The podcasts also seemed very effective in generating useful discussion, both about the content of the discussions and about the use of AI to generate such output.

Structure of the guidelines

Developing this resource is a challenge, not least because the landscape keeps changing so rapidly. To make sure that the guidelines we produce are not just ephemeral in value, we have decided to structure them hierarchically from fixed principles (e.g., “The use of AI should be safe and legal”) to context-specific competence descriptors (“I can evaluate the extent to which AI tools comply with relevant data protection laws in my jurisdiction”, and select only those that respect learner privacy”) and reflection prompts (“Does your school or organisation have a published policy about the use of digital tools and data privacy? Where can you find it? Do students also have access to it?).

Participants in the AI Lang workshop seemed to find this structure both intuitive and empowering. Some suggested that the principles provided a clear anchor for guidance, while the competence descriptors translated those principles into concrete, actionable practices. Others also commented favourably on how the reflection prompts opened space for language teachers teachers to situate AI use within their own institutional realities. The consensus seemed to be that this layered approach made the guidelines feel less like prescriptive rules and more like a scaffold for professional judgment, a way of supporting teachers’ autonomy rather than constraining it.

We found this feedback quite reassuring, but it also pointed us towards areas that need further work. For example, some participants highlighted the need for clearer examples showing how the descriptors might translate into actual classroom decisions. There were also some calls for more specific guidance on how to adapt this framework for contexts with limited digital infrastructure or restrictive institutional policies. These insights will feed directly into the next iteration of the guidelines, which we plan to refine over the coming months.

Rethinking the underlying principles
of AI Lang guidelines

In a previous iteration of the guidelines, we had defined six overarching principles, namely that the use of AI in Language Education should be safe, ethical, purposeful, effective, reflective, and critical. These principles aligned with our vision of what language education should be — in fact, I often said that these are not guidelines for the use of AI in Language Education, but rather Language Education in the age of AI, a subtle but important shift in focus.

Over time, however, our thinking shifted towards a more compact set of principles, as follows:

AI in Language Education should be legal and safe

This principle encodes the non-malfeasance (‘do no harm’) attitude that we believe should be at the heart of every human-centred enterpise. We felt it was important to include both words in the description, since they do not always overlap perfectly. This resonated well with our participants: As one of our participants astutely pointed out “something may be unsafe but legal, in the sense that it has not been prohibited yet; and there are examples of practice that are not legal, but do not pose risk’. Others remarked that ‘legal’ can be a complicated term, unless we can address the underlying question of ‘who decides what is legal?

During the workshop, participants agreed with the importance of the principle, although some raised concerns about whether this was something genuinely language-specific. We plan to address this by making the underlying guidelines, competences and reflection prompts more clearly reflect what is particular about language education (e.g., the fact that many teachers work with vulnerable learners, or the ways in which language learning activities encourage learners to share personal information).

AI in Language Education should be ethical, critical and sustainable

This principle bundles up all the elements that can make AI-assisted language education a force for good. It includes considerations of social justice, linguistic diversity, environmental sustainability, and more. Although the conceptual distinction between this and the previous principle is clear enough, several participants pointed out that this is a distinction that is not always easy to maintain.

Others seemed to feel uncertain about the inclusion of ‘critical’ in this list. This is, perhaps, due to the multiple connotation that the word can have. If one uses ‘critical’ to mean something like ‘creative problem solver’, clearly this doesn’t fit well here. But as used here, ‘critical’ indexes a form of practice in linguistics and education, which looks beyond the immediate functionality of activities and tools and attends to the broader conditions under which they are produced and used. It involves asking probing questions about power, bias, access, ownership and accountability; recognising that technologies are never neutral; and remaining attentive to the ways in which AI systems may reinforce, obscure or disrupt existing inequalities in language education. This is perhaps one area where we need to work harder to clarify what we mean.

AI in Language Education should be effective and purposeful

This principle focuses on the pedagogical application of Artificial Intelligence. ‘Efficient’, in this context, indexes that teachers should use AI in ways that are consistent with current understandings of good practice; ‘purposeful’ that AI adds value to pedagogical activities, by making things possible which would be impossible or prohibitively hard in the past.

For many participants, this was the most substantial part of the guidelines, as it brings to the fore questions about not only whether AI can do something, but whether it should. Discussions quickly moved beyond the novelty of AI tools and towards deeper pedagogical questions: How does AI reshape task design? Which aspects of language learning genuinely benefit from automation or augmentation? Where does AI risk diminishing opportunities for interaction, reflection, or creative struggle, i.e., the moments where learning actually happens?

Teachers also raised important concerns about efficiency being misinterpreted as mere labour-saving. Several participants emphasised that the goal is not to outsource teaching but to use AI in ways that enhance professional judgment, free up cognitive space, or allow learners to do things they could not previously do. Purposefulness, in this sense, requires a clear pedagogical rationale: AI should serve learning, not the other way around.

These conversations revealed a strong appetite for examples that illustrate what “effective and purposeful” AI use looks like in practice. Incorporating such examples will be one of our next priorities as we refine this strand of the guidelines.

AI in Language Education should be reflective and agentive

The thinking behind this principle is teachers and learners (rather than, e.g., Big Data) should be the main decision-makers in AI-assisted language education. Drawing on a rich literature backdrop (e.g., Wallace, 1991; Richards & Lockhart, 1994; Farrell, 2015), we highlighted the need for teachers to engage in reflection in-, on- and for action, and we also discussed how artificial intelligence might constrain or augment agency.

While this is an important principle, it seems that it caused challenges among some workshop participants. One problem was that, unlike other principles, this one proved difficult to apply at the activity level: an AI-assisted language learning task may (or may not) be effective and purposeful, but it’s hard to call it reflective. Also, some participants seemed to find the word ‘agentive’ confusing, due to the terminological similarity with AI agents.

Conclusions and directions forward

On the whole, we felt reassured to note that the Guidelines seem to resonate well with the experience of the experts invited at the AI Lang workshop. The participants agreed that the principles and guidelines we presented were relevant to their needs, and that we did not seem to have left out any points of major concern.

We do, however, also note that there is scope for improvement. For instance, there seems to be consensus among the group that our ‘revised’ structure with four principles, rather than the original six, may have been rather too compact. It is likely that a structure with a larger number of more focused principles might offer greater clarity. We also need to revisit how transparently we communicate how we use key words such as ‘agentive’ and ‘critical’, which carry specific meanings in language (teacher) education. Another ‘blind spot’ that the participants rightly pointed out was that we need to make learners more salient in this guidance, and ensure that the description of teacher-facing processes does not overshadow their experience.

These thoughts will be central in our next iteration of the guidelines, as we move towards the final version that we will submit to the ECML.

What do you think of the way these guidelines are shaping up? If you’d like you can share your thoughts in the comments section or social media (use #AILang to tag us, if you want us to see what you’ve written).

Piloting the AI Lang Moodle course

On the second day of the workshop, our focus was on the professional development course that we have been developing. Just like the previous day, this one also proved quite intense, as the participants and we went through successive sessions of individual exploration and experimentation, group discussions and plenary rounds of feedback.

The feedback from our participants was especially valuable, as our Moodle course is still very much a work in progress. Many users commented favourably on the strong pedagogical focus of this resource, highlighting the ways in which it promotes what we had described as ‘effective’ and ‘purposeful’ language teaching practices. Others commented on the clear, intuitive structure of the course: the division of content into sections like ‘listening’, ‘reading’ etc. helped participants to quickly find the information they needed, ultimately enhancing their overall experience and engagement with the material.

While broadly positive, the feedback also pointed at ways in which we can continue to improve the resource. For example, some suggested that we might want to align our terms with the Companion Volume of the CEFR. Building on that, others proposed adding a section on mediation tasks, and –in fact– have already created a form where teachers can submit examples of how they have been using AI to foster their learners’ mediation skills. Based on another participant’s request, we’re also looking into how to create certificates of completion for those who find such things useful.

If you have examples of good practice, where you use Artificial Intelligence tools to help your learners develop mediation skills, we’d love to hear about them. Scroll down to the ‘how to get involved‘ section to find how to share your ideas with us!

Continuing the conversation

When I talk about our work at the ECML, I often focus on the deliverables on which we are working: the guidelines and the Moodle course. There is a third output, however, which does not appear in our project documentation. It is the community of language education professionals with converging visions, shared concerns and common aspirations, who came together to help us with our work and will hopefully continue to collaborate long after this project has finished.

To make sure that we build on the momentum of this collaboration, we have already set up a LinkedIn Group, where people can share ideas about the use of AI in Language education. We have also created a closed discussion space in the ReaLiTea virtual Community of Practice, which you are all welcome to join, and keep our discussion going (see below for details). If neither of these options seems appealing, do feel very free to reach out in the comment section or contact me by email. I’d love to hear about your experiences and views on how to make AI-assisted language education a catalyst for positive change!

What I take home from this workshop

For me, the AI Lang Workshop was the culmination of an exhausting journey that started many months earlier, when we began work on the project. Bringing together so many remarkable colleagues, with different backgrounds, but a shared vision and purpose, is an accomplishment I feel genuinely proud of. Being able to learn from their insights and having access to their ideas is a privilege.

I am also quite excited about getting back to the guidelines and revising them in light of the feedback we received. Excited because the suggestions were thoughtful, generous and deeply informed by practice; apprehensive because incorporating them will require rethinking some sections almost from the ground up. But this is exactly what makes the process worthwhile: the knowledge that our work will be clearer, more rigorous and more responsive because it shaped collectively rather than in isolation.

In another ECML workshop I recently attended, a participant summed up her feelings in the phrase “I feel that I have found my people.” It was a powerful moment then, and it resonated again in Graz. I felt the same way here—this workshop reminded me that our work is not about ideas, practices, or outputs. Even in a world of AI, it is about real people.