German AI Lab Grubel Raises $3.2M to Give Every Legal Case Its Own AI

Jan de Vries · 2026-09-10

### A Small German Lab With a Big Idea Grubel, a Munich- and Tübingen-based AI research lab, just raised $3.2 million in pre-Seed funding. The goal? Build AI systems that adapt to individual legal matters instead of relying on one-size-fits-all models. The round was led by Point Nine, with angel investors including Jeff Dean (Google's former chief scientist), Chris Ré (Stanford professor and Together AI co-founder), Ion Stoica (UC Berkeley professor and Databricks co-founder), and Harvey co-founders Gabe Pereyra and Winston Weinberg. ### Why Legal AI Has Lagged Behind Here's the thing: complex legal work is messy. It doesn't come with ready-made training data, clear task environments, or obvious tests of success. As Reinhard Heckel, co-founder of Grubel, puts it: "Complex legal matters do not come with ready-made training data, task environments, or clear tests of success. We are automating the AI specialisation loop that identifies the right information, curates it into training data and environments, adapts the system to the matter, and tests whether its work meets the required standard." That's a mouthful, but it points to a real problem. Legal AI has lagged because each case is unique—its facts, documents, terminology, prior work product, and client standards. Today, teams of engineers and lawyers manually build specialised systems for each workflow. It works, but it's slow and expensive. You can't build a separate system for every single legal matter. ### The Specialisation Loop: How Grubel Plans to Fix It Grubel's thesis is simple: every legal matter needs its own AI. To make that happen, they're automating the specialisation process. Their loop has three parts: - **Data engine:** discovers and curates task-relevant data for each matter or client. - **Test-time adaptation layer:** adapts both the model and the agent to the matter before acting. - **Continual evaluation framework:** checks the work against standards derived from the matter itself. The loop repeats until the system delivers what the legal task demands. It's like giving each case its own personal AI assistant that learns on the job. ### The Brains Behind the Operation Grubel was founded in 2026 by machine-learning researchers Moritz Hardt and Reinhard Heckel. Hardt is a director at the Max Planck Institute for Intelligent Systems, a former UC Berkeley professor, and a Google Brain researcher. He co-developed test-time training and co-authored Lawma. Heckel is a TUM professor of machine learning, currently on leave, and formerly a researcher at IBM Research. He focuses on data-centric machine learning and helped create DataComp-LM and OpenThoughts. ### What's Next for Grubel? The company plans to use the capital for research, product development, and team expansion. An initial evaluation showed that Grubel's specialisation loop can improve performance over general-purpose frontier systems on the legal tasks tested. They aim to further develop the technology and expand into other forms of complex knowledge work. As Louis Coppey, Partner at Point Nine, said: "Over the past two to three years, we've seen many legal AI companies emerge, but few are pushing the boundaries of AI research in the legal field as ambitiously as Reinhard, Moritz and the Grubel team. We believe their approach can unlock new capabilities in legal AI and, over time, other forms of complex knowledge work. We're excited to support them on this journey." So keep an eye on Grubel. If they pull this off, it could change how legal work gets done—and maybe even how we think about AI specialisation in other fields.