Lucie: The French Open Source AI for Education, a Premature Launch That Sparks Debate

On January 25, open-source software publisher Linagora abruptly halted the trial version of Lucie, a French open-source artificial intelligence aimed at education. Developed in partnership with the investment plan France 2030 and using the ChatGPT model, this AI aimed to equip students by 2025. However, critical bugs quickly raised questions about the project's maturity.

An ambitious but rushed project

Lucie was born from a call for tenders won by Linagora, as part of a partnership with the France 2030 plan. This initiative aimed to create a French educational solution capable of competing with international AI tools. By focusing on an open-source infrastructure, the project aimed to ensure complete transparency and sovereign control over the data used by the tool.

However, according to the developers themselves, the launch of Lucie was "premature." This rushed launch revealed technical and strategic gaps that compromise the initial goal of preparing for use in schools by 2025.

Bugs that raise questions

During the testing phase, users reported numerous malfunctions, including:

• Basic calculation errors, calling into question the reliability of the responses provided.

• Inconsistencies in the meaning of responses, causing misunderstandings in interactions.

These issues, although common in the initial development of an AI, become concerning in the context of a tool intended for students. Education demands flawless precision and rigor, criteria that Lucie does not yet seem able to meet.

Data collection at the heart of the project

One aspect of the project has also raised questions: the collection of user conversations. This feature, while essential for refining and improving the AI, raises concerns about privacy and data protection, especially in an educational setting. The challenge for Linagora will be to reassure how this data will be anonymized, stored, and used.

An AI not ready for 2025?

The Lucie project presents a real paradox: although ambitious and promising, it highlights the gap between the vision of a sovereign French AI and the reality of technical challenges. With a deadline as close as 2025, it remains to be seen if Linagora can overcome the current obstacles and deliver a reliable and robust solution.

In the meantime, this episode highlights a recurring problem in the AI ecosystem: the discrepancy between expectations and the actual capabilities of tools deployed prematurely. Education, a sector where trust and quality are essential, cannot afford such false starts.

A challenge for French innovation

Despite these difficulties, the Lucie project retains considerable potential. By learning from this episode, Linagora could strengthen the development of its AI and prove that France is capable of producing sovereign, innovative technologies adapted to educational challenges.

It remains to be seen whether the necessary adjustments can be made in time to meet the ambitious goals of the France 2030 plan. One thing is certain: Lucie will need to live up to expectations to establish itself as a key player in the digital education of tomorrow.