Backed by the research of BSC's Language Technologies Unit and the computational power of MareNostrum 5, Galtea is the first company in Spain to provide a technological and scientific basis for the technical evaluation and regulatory compliance of generative AI.
This new technology ensures that AI models are accurate, safe and do not violate current European regulations, as well as significantly reducing the time from technology development to implementation by accelerating the validation phase by 80%.
The irruption of large generative artificial intelligence models could bring a new revolution in different industrial sectors, with considerable impacts on markets and business models due to their significant potential to reduce costs and improve productivity. In Europe alone, an increase of $84 billion in the contribution of AI to GDP is expected between 2024 and 2030, related to a growth in investment and development of AI by companies, which will turn into an annual GDP growth of 15.9% in that period, according to a report by the Statista portal.
However, this new strategic value has not yet turned into real implementation due to the risks involved. Lack of accuracy, model safety and, more recently, concerns about complying with the new European AI regulation are some of the obstacles that have so far prevented a massive market adoption of generative AI systems, which are those that automatically learn to generate unique content for each query in various formats such as text, image, music or code.
In order to address these challenges, the Barcelona Supercomputing Center - Centro Nacional de Supercomputación (BSC-CNS) has launched a new spin off, Galtea, which will provide these companies with the necessary tools to integrate generative AI models into their processes in a safe, reliable and compliant way. This innovative startup is a pioneer in Spain in providing a technological and scientific basis for the technical assessment and regulatory compliance of generative AI.
The basis of Galtea's technology has been developed at BSC by the Language Technologies Unit, which integrates a team of 50 professionals specialized in AI and natural language processing. This unit has extensive experience in the training and evaluation of language models for a variety of tasks and relies on the massive computational power of MareNostrum 5. BSC technology has now been transferred to Galtea, where it will be further refined for commercial applications.
“Our technology, developed by BSC's Language Technologies unit, addresses the main barriers preventing companies from adopting AI: accuracy, security and compliance. We offer solutions that enterprises can rely on,” said Jorge Palomar, CEO of Galtea. “The approach is disruptive because we offer a scientific and systematic methodology to evaluate products based on large language models for specific use cases, adding an integral layer of compliance and governance.”
Barriers to Generative AI adoption
So far, many companies have been reluctant to adopt generative AI models due to three main concerns:
- Lack of accuracy: Many AI models suffer from “hallucinations,” i.e., they generate inaccurate or misleading information, posing significant risks to business applications.
- Safety and security: There are concerns about the appropriate and non-harmful behavior of AI systems, especially in terms of reputational risks, data breaches and biases. This is critical in some industries, such as finance, where biased AI models could unfairly affect loan approvals or hiring decisions.
- Compliance: Adhering to new regulations, such as the EU AI Law, the first comprehensive law on AI worldwide and labeled as 'the most technical law ever', is becoming increasingly necessary for companies developing AI to ensure safety and respect users' fundamental rights.
Galtea is a company that stands out in the market for having the right software infrastructure to provide a solution to these challenges. The combination of the methods and tools developed by BSC's Language Technologies team with cutting-edge scientific infrastructure such as the MareNostrum 5 supercomputer guarantees companies that AI models are accurate, safe and do not violate current European regulations, which provide for fines of up to 35 million euros for non-compliance.
Indispensable solution for deploying AI
BSC's new spin-off offers a comprehensive platform designed for developers and engineers, providing them with the tools they need to integrate generative AI models safely and effectively into their business processes.
The platform includes a dashboard that allows users to monitor and evaluate AI models, providing visibility into their performance, risk and compliance status; plus a Software Development Kit (SDK) and Application Programming Interface (API) that facilitate the seamless integration of Galtea's functionality into existing systems, making it easier for developers to incorporate robust AI governance into their workflows.
The company is currently working with one of Spain's top five banks and a promising AI startup, demonstrating the platform's potential in high-risk environments. Among its main benefits, BSC experts pointed out, are a notable reduction in the time from technology development to implementation, accelerating the validation phase by 80%, and an improvement in product performance of up to 25%.
“Right now it is very easy to develop an AI product, but on the other hand, it is extremely difficult to guarantee that it works as intended. Our solution is indispensable to close that gap,” added Baybars Kulebi, CTO of Galtea. “Our goal is to launch a SaaS (Software as a Service) platform by May 2025 that sets a new standard for AI governance.”
Strategic importance and global leadership
Galtea enters the market at a crucial time. While there are some international companies in the early stages of developing similar solutions, mainly in the U.S., Galtea leads the way thanks to its innovative platform, which bridges the gap between technical assessment and regulatory compliance, bringing legal and developer teams closer together.
Backed by the computing power of MareNostrum 5 and the scientific expertise of BSC, Galtea is positioned to become a global leader in AI governance, helping companies deploy AI solutions that are innovative and compliant with emerging regulations.
“Galtea is not just an extension of BSC's research, it is a pioneering initiative that can help define the future of AI in industries around the world and across diverse domains,” said Marta Villegas, leader of BSC's Language Technologies Unit and Galtea's CSO. “We are excited to bring our expertise and developments to the market and provide companies with the confidence they need to implement AI reliably and securely.”
This article was first published on 16 October by Barcelona Supercomouting Center.
Receive the Funding Newswire each Tuesday, our Policy Bulletin each Thursday, and news about bridging Europe’s east-west innovation gap twice a month in The Widening. 
A unique international forum for public research organisations and companies to connect their external engagement with strategic interests around their R&D system.
