top of page
a new AI architecture
large language models have transformed language and content generation. but they process data point by point, and they cannot capture the structures and dynamics hidden across millions of interconnected data points. terrorist networks, money-laundering operations, defence threats, and disinformation campaigns do not reside in individual documents. they emerge from the patterns between them, and conventional ai was not built to see those patterns. arlequin was built to close that gap, not by building a european version of an existing architecture, but by originating a genuinely new one.
topological neural networks
arlequin is developing proprietary ai models based on topological neural networks, designed to learn from how data is connected, including relationships involving multiple elements at once, capturing coordination cycles and multi-way interactions that conventional models are not designed to efficiently detect. its platform, hudex, analyses relationships across heterogeneous data and delivers traceable, explainable, and auditable results deployable in the air-gapped, sovereignty-sensitive environments governments and defence agencies require. the technology is already deployed with governments and large organisations across western and eastern europe, across counterterrorism, financial crime, judicial investigations, information integrity, and cybersecurity. arlequin was founded by hugo micheron, terrorism specialist and adviser to multiple french heads of state, and antoine jardin, former cnrs research engineer and contributor to france's national ai strategy.
related stories
bottom of page