Arlequin AI, a Paris company building AI models for governments and large organisations, raised €28m (about $32.5m) in a Series A round announced on 10 September 2026. Redalpine and OTB Ventures co-led the round, alongside Bpifrance's Defence Innovation Fund.
Existing investors Vsquared Ventures and 10x Founders increased their stakes, and Xavier Niel and ZEBOX, the start-up accelerator of shipping group CMA CGM, joined. The company said all of the investors were European. It did not disclose a valuation or how much it had raised before this round.
The money will go towards a new type of AI model that Arlequin is developing, based on topological neural networks, as well as hiring and commercial expansion.
“We believe the next advances in AI will come from fundamentally different architectures, not simply from making models larger and feeding them more data and energy.”Antoine Jardin, CTO and co-founder, Arlequin AI
What Arlequin AI does
Arlequin was founded in Paris in 2024 by chief executive Hugo Micheron, who holds a PhD and specialises in jihadism and geopolitical instability, and chief technology officer Antoine Jardin, a former research engineer at France's CNRS working on data science and human behaviour.
The company has built an analytics platform that maps relationships across mixed data such as documents, financial transactions, video and operational records, with each result traceable to the underlying evidence. In one example given by the company, the platform analyses millions of data points from devices seized in a counterterrorism investigation to link people, places, communications and events. Arlequin says the technology is used by governments and large organisations in Western and Eastern Europe and is active in four European countries. Its target uses include security and defence, criminal and fraud investigations, anti-money-laundering work, information integrity and cybersecurity.
The company has about 50 employees, including 35 engineers and 15 PhDs, researchers and data scientists. It is headquartered in Paris and has offices in London and Berlin.
A different model architecture
The new funding is meant to pay for Arlequin's own models based on topological neural networks, which are designed to learn from how data points are connected, including relationships involving several elements at once, rather than from individual data points alone. The company presents this as a different path from the large language models that have dominated recent AI development.
Arlequin says the architecture is being designed to need much less computing power than large AI systems, reducing reliance on energy, advanced chips and data-centre capacity. RuntimeWire noted that these efficiency claims had not yet been tested in public benchmarks. The company said the work is supported by research collaborations with INRIA, CNRS, the Max Planck Institute and the universities of Oxford, Cornell, Princeton and UC Santa Barbara.
What the money is for
Arlequin said the round would let it expand the international team that develops and trains its models and speed up commercial deployment in Europe and elsewhere. It plans to open an AI lab in Silicon Valley by the end of 2026.
Oliver Pabst, a general partner at redalpine, said the founders had a strong fit with their market and were building a sovereign security platform for Europe. Maria Portoles of Vsquared said the company combined fundamental research with a product already used by governments and enterprises. Micheron framed the round in terms of Europe's push for technological independence.
Company website: arlq.ai