Galtea secures $3.2 million to assist businesses in evaluating AI agents.

      The spin-off from the Barcelona Supercomputing Center, established eighteen months ago, leverages AI to create realistic testing scenarios that reveal failures, hallucinations, biases, and security risks before deploying enterprise AI agents. The funding round was led by 42CAP, with participation from Mozilla Ventures.

      The distinction between an AI agent that functions well in a demo and one that operates effectively in a production environment increasingly hinges on testing. Developing a generative AI product is no longer the challenging aspect; the necessary models are available, APIs are easy to access, and development tools have advanced.

      However, the infrastructure needed to ensure that what you’ve built will function reliably when real users engage with it at scale has not kept pace. Galtea, a startup based in Barcelona and spun out of the Barcelona Supercomputing Center in October 2024, is addressing this need. The company has secured $3.2 million in seed funding to further its efforts.

      42CAP, an early-stage technology fund based in Munich, led this funding round, with contributions from Mozilla Ventures, which focuses on trustworthy AI, as well as existing investors JME Ventures, Masia, and ABAC Nest Ventures. The total funding now amounts to $4.1 million. The fresh capital will be used to enhance Galtea’s engineering and commercial teams and to further develop the platform.

      Galtea’s primary product generates test cases and synthetic user simulations based on the intended behavior of an AI agent. The objective is to automatically create adversarial and edge-case scenarios at scale, eliminating the need for engineering teams to manually write these scenarios, a process that is often lengthy, costly, and incomplete.

      The platform assesses models concerning hallucination rates, bias, security vulnerabilities, and toxicity, providing structured metrics for developers and compliance teams to inform deployment decisions. This week, the company also introduced a self-service tier with a free trial, expanding access beyond its current enterprise clientele.

      The founders bring a unique background to the industry. CEO Jorge Palomar previously worked at Amazon and within BSC’s Language Technologies research group before founding Galtea. His co-founder, CTO Baybars Külebi, possesses a PhD in Astrophysics, has co-founded several language and audio technology initiatives, and has extensive experience as a machine learning specialist at BSC.

      The technology originated at BSC for the evaluation of large language models for internal research, utilizing MareNostrum 5, one of Europe’s most powerful supercomputers. This background lends Galtea a significant scientific depth for a company of its age.

      The timing for commercial use is strategic, as the EU AI Act now mandates that companies deploying AI in high-risk applications document and validate their models’ safety and compliance, with potential fines reaching up to €35 million for noncompliance.

      For European companies that have been creating AI products without a systematic testing infrastructure, the regulation has intensified urgency. Galtea’s platform directly addresses this gap, assisting development and legal teams in generating the necessary compliance evidence without the need to overhaul existing workflows.

      In particular, Mozilla Ventures' participation indicates that this funding round was partially framed around the trustworthy-AI narrative that has shaped the fund’s focus since its inception in 2022.