AI company Textgain is set to build what it claims is the world’s first comprehensive Large Language Model to detect hate speech in all official European languages. The Belgian firm, founded as an AI spin-off from the University of Antwerp’s Computational Linguistics & Psycholinguistics Research Center, will develop its LLM, named CaLICO, over the next […]
AI company Textgain is set to build what it claims is the world’s first comprehensive Large Language Model to detect hate speech in all official European languages.
The Belgian firm, founded as an AI spin-off from the University of Antwerp’s Computational Linguistics & Psycholinguistics Research Center, will develop its LLM, named CaLICO, over the next 12 months, promising new standards for multilingual AI-based hate speech detection.
The firm was recently named one of four winners of the European Commission’s Large AI Grand Challenge, which saw 94 international AI developers compete for a share of a €1 million prize.
Textgain was named victor alongside French AI fintech Lingua Custodia, Latvian chatbot firm Tilde, and translation firm Unbabel, whose CEO Vasco Pedro spoke with TechInformed at Web Summit in Lisbon last year.
Textgain was awarded €250,000 in the competition led by AI-BOOST. The firm says it will develop its own language model to ensure high reliability and precision, which are essential for effective content moderation.
CEO Guy De Pauw highlighted this type of project’s unique challenges and solutions: “Large language models, especially commercial ones, refuse to process toxic language. This makes it almost impossible to use them to process hate speech. We are now building a language model from scratch that can process this type of content — but without generating it. That’s what makes us different.”
Winners were also rewarded with a share of 8 million GPU hours on two of the world-leading EuroHPC JU supercomputers, LUMI and LEONARDO.
According to COO Redouan el Hamouchi, access to the supercomputer will significantly accelerate the development of their new AI model: “We are pleased that we can train our application faster so that it can handle different languages and cultures within the European Union.”
Textgain plans to leverage its AI capabilities for international expansion. The company’s focus for the coming year will be developing the CaLICO model, which is valued at several million euros.
Beyond that, Textgain aims to develop SaaS applications, positioning itself as a market leader in AI-based hate speech detection.
De Pauw elaborated on its strategic vision: “We also distinguish ourselves through our academic approach. We work together with policymakers, security services, social organisations and scientists. Reliability and precision are important rather than generic solutions and unrealistic promises.”
Textgain added that it was committed to ethical AI development. According to EU AI policy, “The technology must be transparent, explainable, and ethical,” said De Pauw. “These values are also in our DNA. It’s the only way to give people confidence in new technology.”
With elections taking place across the globe in 2024, disinformation on social media, the tech could play a significant role in this year of elections.
