From interviewing terrorists to rethinking AI: Arlequin raises €28M for Europe’s alternative to LLMs
Tech Funding News Sofia Chesnokova ● Covered by 2 sources
Arlequin AI raised €28M to push an AI system that doesn’t rely on LLMs. It’s betting Europe can build auditable AI instead of chasing bigger chatbots.
Based on reporting by Tech Funding News, Sofia Chesnokova — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Paris-based Arlequin AI has landed a €28 million Series A to scale a very specific bet: an AI architecture that does not depend on large language models. The round was led by redalpine and OTB Ventures, with Bpifrance’s Defence Innovation Fund also in the mix. Xavier Niel came in as a new investor, while Vsquared Ventures and 10x Founders increased their stakes.
That brings Arlequin’s total funding to €32.4 million. The company, founded two years ago by Hugo Micheron and Antoine Jardin, now says it has about 50 employees, including at least 35 engineers and 15 PhDs and postdocs. It also claims more than 30 clients across four European countries.
Micheron’s route into AI is unusual even by startup standards. He learned Arabic in Syria, did extensive fieldwork on European jihadism, and interviewed hundreds of convicted terrorists as part of his academic research. Later, while lecturing at Princeton from 2020 to 2023, he became frustrated by the AI tools he saw as biased, and that pushed him toward building Arlequin.
The company’s pitch is that language models are fine for conversation, but not for analysis where mistakes can have real consequences. Arlequin uses an LLM only as a front end for questions; the actual work is done by unsupervised models that look for relationships in video, audio, text, images and seized devices, with results meant to be traced and audited. Micheron says the core architecture, a topological neural network, is meant to focus on connections in data rather than language patterns, and that only about 20 researchers worldwide are working in this area.
Arlequin is also dressing the raise in geopolitical terms. It says the US has the computing and data advantage, China is not a model Europe wants to copy, and Europe’s Mistral effort is still smaller in scale. The startup plans to keep growing in France, move into Germany and open a London office, with the first employees there starting this week. Its valuation hasn’t been disclosed.
My take — AI-written commentary, not fact-checked reporting
This is the rare AI startup pitch that sounds less like hype and more like a refusal to join the chatbot parade. Europe keeps saying it wants sovereignty, and here’s a team trying to build something it can inspect without crossing its fingers. That alone makes the bet more interesting than another larger model with a shinier demo.
Read more about this at: Tech Funding News