Exponential Security Labs

Exponential Security Labs GmbH, based in Tübingen, Germany, builds multi-agent security environments: complex scenarios with live services, persistent state and verifiable outcomes. The site says these environments are used for adversarial training, where attacker and defender agents are trained against one another, and for adaptive benchmarking of how other AI agents operate under realistic adversarial conditions.

The website does not explicitly name its customers or describe its revenue model. The company presents its team as working on AI safety and security for nearly a decade, with more than 20 papers at leading AI conferences over the last five years and contributions to the International AI Safety Report and to benchmarks such as AgentHarm, OS-Harm, JailbreakBench and RobustBench. Naman Deep Singh and Anselm Paulus are listed as co-founders, Maksym Andriushchenko and Matthias Hein as founding advisors, and the company is described as backed by Mätch VC.

The imprint lists Exponential Security Labs GmbH, registered at the Amtsgericht Stuttgart, with its address in Tübingen. The careers page states the company is hiring in Tübingen on a hybrid setup and lists five open roles: two Member of Technical Staff positions, a Research Intern position, a Chief of Staff role and an open application. The founding year is not stated on the website.

Product: Multi-agent security environments for reinforcement learning training, adversarial training of attacker and defender agents, and adaptive benchmarking of AI agents.