Sherpa.ai from Spain secures $18 million for data-sovereign artificial intelligence.

      Many are eager to leverage the power of AI, but few are willing to entrust their most sensitive information to a foreign cloud provider. A Basque startup believes it has the solution.

      Sherpa.ai has secured $18 million to develop AI that does not access raw data, as reported by tech.eu. This Spanish company caters to banks, hospitals, and government entities, which are particularly concerned about data security. Its approach aligns with an emerging concept: sovereign AI.

      Training without data sharing

      At the core of Sherpa.ai's technology is federated learning. Rather than centralizing data for model training, the model is sent to the data. Each bank or hospital trains the model locally and only shares the insights, not the actual records. Sherpa.ai claims that this method can reduce the data exchanged between locations by as much as 99 percent.

      This is especially significant in regulated industries where privacy laws hinder typical AI initiatives. According to founder and CEO Xabi Uribe-Etxebarria, the objective is to enable organizations to "maximize the advantages of AI without losing control, privacy, and sovereignty over their data."

      A European approach with American support

      The latest updates from the EU tech sector include insights from our founder Boris and some intriguing AI-generated art. Subscribe now for weekly updates directly to your inbox! Forgepoint Capital, an investor in cybersecurity and AI from Silicon Valley, participated in this funding round, alongside existing investors Mundi Ventures, Ekarpen, Allegra Holdings, and SETT. This reflects a notable American endorsement for a company addressing the very concerns that are prompting Europe to prioritize in-house AI solutions.

      The client roster also indicates this trend. Sherpa.ai recently signed contracts with Spain's Indra, banks Caja Laboral and Unicaja, security firm Prosegur, genomics company Centogene, and the US National Institutes of Health. A European firm prioritizing privacy while working with a US federal agency serves as compelling evidence of its approach.

      Substance behind the concept

      "Sovereign AI" is a widely used term, and federated learning is not a new concept. What lends credibility to Sherpa.ai's version is the research backing it. The company has published peer-reviewed studies on training large language models using private datasets and collaborated with the NIH and University College London on employing this technique for diagnosing rare diseases.

      While the funding is relatively modest and the market is competitive, there are numerous companies claiming to offer AI that respects data sovereignty, ranging from national projects to privacy-centric security startups. However, the demand is substantial and increasing. As governments impose stricter regulations on data storage, the company that can train models without accessing the data has a compelling narrative.

      Sherpa.ai is banking on the belief that this narrative will resonate, backed by its $18 million investment.