Cloudflare Adds EuroLLM, Apertus, and Open Security Harness to Workers AI
Cloudflare announced the addition of two major multilingual open-source LLMs, EuroLLM and Apertus, to Workers AI, along with an open-source model-agnostic AI security harness. These changes improve language support, regulatory compliance, and security flexibility for developers.
What changed?
Cloudflare announced that EuroLLM and Apertus, two large-scale, multilingual, open-source language models developed by European and Swiss research institutions, are now available (by request) through Workers AI. EuroLLM supports 35 languages, including all 24 official EU languages, while Apertus covers more than 1,500 languages, with a substantial portion of its training data in non-English languages. Apertus was also built with EU AI Act and GDPR compliance in mind.
Additionally, Cloudflare released and open-sourced a security harness orchestration layer. This harness can coordinate multiple AI models—open or closed—in parallel to discover vulnerabilities, verify findings, and prioritize security threats. The harness is model-agnostic, removing dependence on a single provider.
Why does it matter to an everyday developer?
These changes expand language and region support for developers building multilingual or globally accessible apps. EuroLLM and Apertus provide out-of-the-box support for dozens to thousands of languages, including those underserved by other models, making it easier to reach users in regulatory-sensitive or linguistically diverse regions such as Europe and Africa.
The security harness means developers (especially in security-sensitive, government, or infrastructure roles) can orchestrate AI-powered security checks without vendor lock-in. If access to one AI model is restricted or lost, defenses can continue by swapping in alternative models, reducing operational risk.
What can the developer do now?
- Request access to EuroLLM and Apertus through Workers AI to experiment or build products with advanced multilingual capabilities.
