Cloudflare Adds EuroLLM and Apertus to Workers AI, Launches Cyber Defense Workshops

Cloudflare is expanding Workers AI with two new open models built by European public institutions: EuroLLM and Apertus. This move reinforces the company's argument that AI sovereignty relies on choice and open access rather than restricted, single-provider dependencies.

New European Open Models

EuroLLM supports 35 languages, including all 24 official EU languages, and was developed by a consortium that includes Instituto Superior Técnico and the University of Edinburgh. Trained on the MareNostrum 5 supercomputer, it reportedly outperforms similar-sized models on EU multilingual benchmarks and machine translation tasks.

Apertus, Switzerland's first fully open model, was trained on over 1,500 languages with 40% of its data in non-English languages. Developed by ETH Zurich, EPFL, and the Swiss National Supercomputing Centre (CSCS), it was built on the Alps supercomputer and designed to comply with the EU AI Act and GDPR. Its architecture, weights, and training data are fully public.

Cybersecurity Workshops and Open-Source Tools

Cloudflare is also launching hands-on workshops for government cyber agencies and critical infrastructure operators to help them build AI defenses that work with any model. The first workshop takes place in Singapore this October during Singapore International Cyber Week.

These workshops are based on Cloudflare's open-sourced AI security harness, a system that coordinates multiple AI models to find and verify vulnerabilities. By using an open-source harness, organizations can avoid vendor lock-in and maintain security even if access to a specific frontier model is restricted.