TrustLLM

Happy to share that our paper, “Not All Secrets Are Equal: Type-Aware Unlearning for Language Model Secret Removal,” has been accepted to ECML PKDD 2026 🎉 In this work, we study how to surgically remove memorized secrets from language models without affecting the model’s broader capabilities. We find that different types of secrets behave differently during unlearning, which led us to PSTU. Looking forward to presenting this at ECML PKDD 2026:  (https://ecmlpkdd.org/2026/) Big thanks to Amath SOW and Andreas Bueff, and to my supervisors Jiahui Geng, Mattias Tiger, and Fredrik Heintz for their continued guidance, valuable feedback, and collaboration throughout this project. #ECMLPKDD #MachineLearning #LLM #AIResearch This work was supported by TrustLLM and funding from the European Union 🇪🇺

𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 𝐞𝐯𝐞𝐧𝐭 - GeTT-together with Europe: 📍 Cologne | 21 May, 09:00 CET | AI Village, Fraunhofer room Looking forward to an exciting public event in Cologne, hosted by TrustLLM, as part of the 𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 (𝐆𝐞𝐓𝐓) – 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫 𝐰𝐢𝐭𝐡 𝐄𝐮𝐫𝐨𝐩𝐞 𝐢𝐧𝐢𝐭𝐢𝐚𝐭𝐢𝐯𝐞. At AI Village on 21 May, 09:00 CET, we will bring together researchers, industry leaders and innovators to discuss how large language models are being implemented responsibly and effectively across European industry and society. (𝐍𝐨𝐭𝐞 𝐭𝐡𝐚𝐭 𝐭𝐡𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 𝐦𝐚𝐲 𝐛𝐞 𝐮𝐩𝐝𝐚𝐭𝐞𝐝) 🗓 𝐓𝐞𝐧𝐭𝐚𝐭𝐢𝐯𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 09:00 – 09:30 𝐅𝐫𝐞𝐝𝐫𝐢𝐤 𝐇𝐞𝐢𝐧𝐭𝐳 – 𝐏𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 09:30 – 10:00 𝐃𝐞𝐧𝐳𝐢𝐥 𝐉𝐨𝐲 (𝐓𝐍𝐎) – 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 𝐮𝐬𝐞 𝐜𝐚𝐬𝐞𝐬 𝐚𝐧𝐝 𝐭𝐞𝐜𝐡 𝐭𝐫𝐚𝐧𝐬𝐟𝐞𝐫 10:00 – 11:30 | 𝐏𝐚𝐧𝐞𝐥 𝐝𝐢𝐬𝐜𝐮𝐬𝐬𝐢𝐨𝐧 Moderator: Valentina Kerst 𝐏𝐚𝐧𝐞𝐥𝐥𝐢𝐬𝐭𝐬: • Fredrik Heintz, Professor at Linköping University • Lara Lawniczac, Soofi • Lucie Flek, Professor, University of Bonn • Stephan Oepen, Professor University of Oslo The session will focus on trustworthy AI, cross-border collaboration and practical pathways from research to deployment. If you’re interested in the future of LLMs in Europe, from policy and research to real-world impact, this is a conversation worth joining. 11:30 – 12:00 Nicolas Flores-Herr, Fraunhofer IAIS More information 👉 https://lnkd.in/dx9t66XJ

𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 𝐞𝐯𝐞𝐧𝐭 - GeTT-together with Europe: 📍 Cologne | 21 May, 09:00 CET | AI Village, Fraunhofer room Looking forward to an exciting public event in Cologne, hosted by TrustLLM, as part of the 𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 (𝐆𝐞𝐓𝐓) – 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫 𝐰𝐢𝐭𝐡 𝐄𝐮𝐫𝐨𝐩𝐞 𝐢𝐧𝐢𝐭𝐢𝐚𝐭𝐢𝐯𝐞. At AI Village on 21 May, 09:00 CET, we will bring together researchers, industry leaders and innovators to discuss how large language models are being implemented responsibly and effectively across European industry and society. (𝐍𝐨𝐭𝐞 𝐭𝐡𝐚𝐭 𝐭𝐡𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 𝐦𝐚𝐲 𝐛𝐞 𝐮𝐩𝐝𝐚𝐭𝐞𝐝) 🗓 𝐓𝐞𝐧𝐭𝐚𝐭𝐢𝐯𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 09:00 – 09:30 𝐅𝐫𝐞𝐝𝐫𝐢𝐤 𝐇𝐞𝐢𝐧𝐭𝐳 – 𝐏𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 09:30 – 10:00 𝐃𝐞𝐧𝐳𝐢𝐥 𝐉𝐨𝐲 (𝐓𝐍𝐎) – 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 𝐮𝐬𝐞 𝐜𝐚𝐬𝐞𝐬 𝐚𝐧𝐝 𝐭𝐞𝐜𝐡 𝐭𝐫𝐚𝐧𝐬𝐟𝐞𝐫 10:00 – 11:30 | 𝐏𝐚𝐧𝐞𝐥 𝐝𝐢𝐬𝐜𝐮𝐬𝐬𝐢𝐨𝐧 Moderator: Valentina Kerst 𝐏𝐚𝐧𝐞𝐥𝐥𝐢𝐬𝐭𝐬: • Fredrik Heintz, Professor at Linköping University • Lara Lawniczac, Soofi • Lucie Flek, Professor, University of Bonn • Stephan Oepen, Professor University of Oslo The session will focus on trustworthy AI, cross-border collaboration and practical pathways from research to deployment. If you’re interested in the future of LLMs in Europe, from policy and research to real-world impact, this is a conversation worth joining. 11:30 – 12:00 Nicolas Flores-Herr, Fraunhofer IAIS More information 👉 https://lnkd.in/dx9t66XJ

𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 𝐞𝐯𝐞𝐧𝐭 - GeTT-together with Europe: 📍 Cologne | 21 May, 09:00 CET | AI Village, Fraunhofer room Looking forward to an exciting public event in Cologne, hosted by TrustLLM, as part of the 𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 (𝐆𝐞𝐓𝐓) – 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫 𝐰𝐢𝐭𝐡 𝐄𝐮𝐫𝐨𝐩𝐞 𝐢𝐧𝐢𝐭𝐢𝐚𝐭𝐢𝐯𝐞. At AI Village on 21 May, 09:00 CET, we will bring together researchers, industry leaders and innovators to discuss how large language models are being implemented responsibly and effectively across European industry and society. (𝐍𝐨𝐭𝐞 𝐭𝐡𝐚𝐭 𝐭𝐡𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 𝐦𝐚𝐲 𝐛𝐞 𝐮𝐩𝐝𝐚𝐭𝐞𝐝) 🗓 𝐓𝐞𝐧𝐭𝐚𝐭𝐢𝐯𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 09:00 – 09:30 𝐅𝐫𝐞𝐝𝐫𝐢𝐤 𝐇𝐞𝐢𝐧𝐭𝐳 – 𝐏𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 09:30 – 10:00 𝐃𝐞𝐧𝐳𝐢𝐥 𝐉𝐨𝐲 (𝐓𝐍𝐎) – 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 𝐮𝐬𝐞 𝐜𝐚𝐬𝐞𝐬 𝐚𝐧𝐝 𝐭𝐞𝐜𝐡 𝐭𝐫𝐚𝐧𝐬𝐟𝐞𝐫 10:00 – 11:30 | 𝐏𝐚𝐧𝐞𝐥 𝐝𝐢𝐬𝐜𝐮𝐬𝐬𝐢𝐨𝐧 Moderator: Valentina Kerst 𝐏𝐚𝐧𝐞𝐥𝐥𝐢𝐬𝐭𝐬: • Fredrik Heintz, Professor at Linköping University • Lara Lawniczac, Soofi • Lucie Flek, Professor, University of Bonn • Stephan Oepen, Professor University of Oslo The session will focus on trustworthy AI, cross-border collaboration and practical pathways from research to deployment. If you’re interested in the future of LLMs in Europe, from policy and research to real-world impact, this is a conversation worth joining. 11:30 – 12:00 Nicolas Flores-Herr, Fraunhofer IAIS More information 👉 https://lnkd.in/dx9t66XJ

𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 𝐞𝐯𝐞𝐧𝐭 - GeTT-together with Europe: 📍 Cologne | 21 May, 09:00 CET | AI Village, Fraunhofer room Looking forward to an exciting public event in Cologne, hosted by TrustLLM, as part of the 𝐆𝐞𝐫𝐦𝐚𝐧𝐢𝐜 𝐓𝐞𝐜𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐞𝐫 (𝐆𝐞𝐓𝐓) – 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫 𝐰𝐢𝐭𝐡 𝐄𝐮𝐫𝐨𝐩𝐞 𝐢𝐧𝐢𝐭𝐢𝐚𝐭𝐢𝐯𝐞. At AI Village on 21 May, 09:00 CET, we will bring together researchers, industry leaders and innovators to discuss how large language models are being implemented responsibly and effectively across European industry and society. (𝐍𝐨𝐭𝐞 𝐭𝐡𝐚𝐭 𝐭𝐡𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 𝐦𝐚𝐲 𝐛𝐞 𝐮𝐩𝐝𝐚𝐭𝐞𝐝) 🗓 𝐓𝐞𝐧𝐭𝐚𝐭𝐢𝐯𝐞 𝐚𝐠𝐞𝐧𝐝𝐚 09:00 – 09:30 𝐅𝐫𝐞𝐝𝐫𝐢𝐤 𝐇𝐞𝐢𝐧𝐭𝐳 – 𝐏𝐫𝐞𝐬𝐞𝐧𝐭𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 09:30 – 10:00 𝐃𝐞𝐧𝐳𝐢𝐥 𝐉𝐨𝐲 (𝐓𝐍𝐎) – 𝐓𝐫𝐮𝐬𝐭𝐋𝐋𝐌 𝐮𝐬𝐞 𝐜𝐚𝐬𝐞𝐬 𝐚𝐧𝐝 𝐭𝐞𝐜𝐡 𝐭𝐫𝐚𝐧𝐬𝐟𝐞𝐫 10:00 – 11:30 | 𝐏𝐚𝐧𝐞𝐥 𝐝𝐢𝐬𝐜𝐮𝐬𝐬𝐢𝐨𝐧 Moderator: Valentina Kerst 𝐏𝐚𝐧𝐞𝐥𝐥𝐢𝐬𝐭𝐬: • Fredrik Heintz, Professor at Linköping University • Lara Lawniczac, Soofi • Lucie Flek, Professor, University of Bonn • Stephan Oepen, Professor University of Oslo The session will focus on trustworthy AI, cross-border collaboration and practical pathways from research to deployment. If you’re interested in the future of LLMs in Europe, from policy and research to real-world impact, this is a conversation worth joining. 11:30 – 12:00 Nicolas Flores-Herr, Fraunhofer IAIS More information 👉 https://lnkd.in/dx9t66XJ

OSFM 2.0 Workshop: Open Source Foundation Models 21–22 May 2026🚀 We are pleased to share the current agenda for our OSFM workshop, which focuses on the topics that are currently shaping the European AI landscape. Thank you to all participants for their valuable input and suggestions, which have helped to shape the program. The workshop is divided into three thematic sections: large language models, policy and governance, and industry and infrastructure. A key cross-cutting theme throughout is AI sovereignty, framing discussions across all sessions. Besides interactive formats and discussions, we are looking forward to informative talks of Nicolas Flores Herr, Joerg Bienert, Djamé Seddah, Jan Hajic, Fredrik Heintz,, Hans Uszkoreit, Axel Voss, Joshua Tan, Ilja Rausch, Patrick Putzky, Peter Koerner, Daniel Weingaertner, Stefan Bergstein, Thomas Feld, Joern Kellermann.   Interested in joining? A limited number of seats are still available. Please register here: https://lnkd.in/eqMvE7VB More information: osfm.info/agenda