A Guide to the Framework for Artificial Intelligence
Over the past few years, Artificial Intelligence (AI) tools have been developed and gradually made available to the general public. The launch of the ChatGPT tool by OpenAI in late 2022 met with considerable success, extending well beyond the communities of computer scientists and even tech enthusiasts, particularly amongst the younger generations.
The availability of these tools and their ease of use undeniably has a considerable impact on all higher education and research institutions and, more broadly, on society. INALCO is particularly affected, as these tools offer text generation (‘genAI’) capabilities, the quality of which, however, varies greatly depending on the language and is questionable for a number of uses.
This guidance document, written for Inalco staff (academic staff and administrative staff) and students, aims primarily to explain the features offered by AI tools, to describe use cases in various contexts (teaching, administrative tasks, research work), to provide recommendations and to set out the rules that must be followed when using these tools. It does not seek to take a general stance on the impact of these tools on society.
Definitions
Artificial Intelligence
Algorithms falling within the field of Artificial Intelligence (AI) aim to replicate human cognitive abilities through calculations that can be performed on an electronic device (such as computers, tablets or mobile phones).
Generative AI
Generative AI models use algorithms and models to determine the most likely words in a process of artificially generating data (text, images, videos, etc.). Generation is initiated by a ‘prompt’ (a command written in natural language). The range of use cases is broad.
Large language models
Large language models (LLMs) are computational models constructed by fine-tuning a vast number of parameters calculated through ‘machine learning’, based on very large volumes of text. They model language by identifying patterns in textual data. Their parameters and statistical calculations enable them, amongst other things, to ‘generate’ coherent texts by simulating human abilities related to the manipulation of text and underlying knowledge. These models can be developed for a single language or for several languages simultaneously, depending on the textual data used to train them. The performance of these models varies significantly across languages; generally, languages that are widely used on the internet (English, French, Spanish) yield better results than languages with limited digital resources.
Uses of AI
General-purpose AI systems offer and enable the performance of numerous tasks, including the following non-exhaustive list:
- information retrieval, such as a search engine,
- identifying issues and searching for references,
- organising ideas,
- checking a text and suggesting improvements,
- summarising texts or documents,
- automatic translation of texts,
- digitisation or transcription (of images for text or audio files for speech).
It is worth noting that this list includes many of the fundamental skills that school and university education aims to help students acquire. Furthermore, it should be borne in mind that the use of AI can hinder learning, or even lead to a loss of skills or a dependence on tools. Therefore, unless AI is itself part of the learning objectives, it is strongly recommended that it be avoided so that the assistance provided does not replace the learning process itself.
Furthermore, in cases where the use of AI is permitted or envisaged by the institution or the teacher, it is essential to be aware of the potential benefits and risks of each application and, if its use is not problematic, to determine the methodology to be adopted to properly frame the practice. The quality of the output remains, in fact, highly variable depending on several factors (task, language, input data, etc.). There may also be specialised AI-based tools for each use case; we do not currently provide a list of these, as there are numerous tools available and they are evolving very rapidly.
Finally, it is important to be aware that these AI tools are gradually being integrated into existing applications where the use of AI is not always explicitly mentioned (search engines, spell-checkers in word processors, summaries for document reading, transcriptions of video conferences, etc.)
Best practice recommendations
When using AI tools, it is essential to scrutinise the output produced by the tool with a critical eye and a great deal of vigilance. A very large number of cases of ‘hallucinations’ have been identified (where the AI makes claims that are not based on any verifiable source), particularly in specialist fields and in the generation of bibliographic references. Where AI is used in a piece of work, it is recommended to state this explicitly and to indicate which parts of the work involved the use of AI tools (data analysis, information retrieval, translation, text refinement, etc.).
If the use of AI is being considered, it is strongly recommended that it be used to provide assistance and suggestions, and not to generate the work itself. In this context, an AI system may offer comments on a piece of academic work, which the author may review and, following verification, potentially take into account when incorporating them into the work in question.
We urge those using AI to be particularly vigilant regarding the potentially biased or incomplete nature of the outputs. These outputs are, in fact, designed to provide answers based on sources selected by AI providers, which are often partial and incomplete, and very rarely admit to their own limitations or lack of knowledge, particularly in fields requiring specialist expertise. Furthermore, their use in the context of research carries the risk of limiting the work to the avenues suggested by the AI; it is therefore strongly recommended not to use them, or to use them sparingly and with discernment, when initiating research projects.
Furthermore, it is common knowledge that AI systems are extremely energy-intensive; their use and the deployment of their functionalities must therefore be carefully considered and, as far as possible, weighed against the energy consumption they entail. Their use must remain as sparing as possible and, in any event, restricted to the cases for which they are intended and permitted.
Finally, the use of AI in the context of activities at INALCO must, as far as possible, be authorised. You may contact the AI team (link below) if you notice any inappropriate encouragement to use AI, in order to report it and raise an objection to the use in question.
Prohibitions
In the context of activities carried out at INALCO, we ask you to observe the following four prohibitions:
- do not upload personal or confidential data to an online AI tool,
- do not claim an AI-generated text as your own work,
- do not submit material subject to intellectual property rights or copyright (articles, course materials) that has not been made public without the prior authorisation of the author(s),
- do not use machine translation in language assignments, unless its use is explicitly authorised by the tutor.
Please note that the use of AI is currently explicitly prohibited in Inalco’s internal regulations (Article 71) as well as in the regulations governing assessment procedures (Article 2.2). Consequently, any use of AI that contravenes the prohibitions set out above is liable to disciplinary action. We are currently drafting a document that will set out the permitted and prohibited uses of AI within the institution.
Page updated on 20 July 2026.