Cloudflare has added two European public AI models, EuroLLM and Apertus, to its Workers AI platform. Developers can request access to both open models, which were created by research consortia and public institutions to address regional language requirements and regulatory standards.
European Open Models Join Workers AI
EuroLLM was built by a European consortium including Instituto Superior Técnico, the University of Edinburgh, and Unbabel, with support from Horizon Europe and EuroHPC. Trained on the MareNostrum 5 supercomputer, it supports 35 languages, including all 24 official EU languages. According to the consortium, it outperforms similarly sized models on EU multilingual and translation benchmarks.
Apertus is a fully open multilingual model developed by ETH Zurich, EPFL, and the Swiss National Supercomputing Centre on the Alps supercomputer using over 10,000 GH200 GPUs. It was trained on more than 15 trillion tokens across 1,500 languages, with 40 percent non-English data. Its developers published its architecture, weights, training data, and methodology while designing it to align with GDPR and EU AI Act requirements, such as handling opt-outs and data removal. Its creators report strong performance on regional languages like Romansh and Swiss German alongside low-resource global languages.
Model-Agnostic Cyber Defenses
Cloudflare also announced hands-on training workshops for government cybersecurity agencies and operators of critical infrastructure. The curriculum focuses on building model-agnostic AI defenses using an orchestration layer harness that runs multiple models in parallel to analyze threats. Because the harness supports both open and closed models, organizations can maintain defensive systems even if access to a specific provider is interrupted.
The security harness has been open-sourced for independent deployment. Cloudflare's first hands-on workshop is scheduled for October 2026 in Singapore during Singapore International Cyber Week.
AI-generated editorial illustration; not a documentary photograph.
Source: One year later: Sovereign AI and the fight for choice
SIGNAL TITANS
Leave a Reply