Breaking cybersecurity news, news analysis, commentary, and other content from around the world, with an initial focus on the Middle East & Africa, the Asia Pacific, Europe, and Latin America. Europe's Multilingual Reality Exposes AI Security Gaps The AI security layer and guardrails for many AI products don't evenly protect against jailbreaking and unsafe actions in every single language. Not all languages are treated equally when it comes to AI model function and safety, and European organizations face a particular risk when it comes to this reality. The modern large language model (LLM) ecosystem relies heavily on natural language, whether a user is speaking to a chatbot, issuing specific instructions for software development, generating emails, or performing large-scale data analysis. This reliance is further illustrated through the wide range of prompt injection attacks that rely on language-based trickery. While leading models can process text in dozens to hundreds of languages, performance and safety capabilities vary dramatically between languages. Many of the less-supported languages, such as Welsh and Swahili, can answer only basic prompts and may make grammatical errors. Mainstream models like OpenAI's GPT, Google's Gemini, and Anthropic's Claude demonstrate strong performance in around 30 to 40 languages such as English, Arabic, Spanish, French, German, Japanese, Simplified Chinese, and Hindi. English is by far the language best supported by many AI models. English benefits both from disproportionate training data and from tokenization schemes that often represent English more efficiently than many other languages. Academic research shows that many LLMs perform logic, reasoning, coding, and math tasks best when prompted in English language, and many major AI labs conduct safety tuning, behavior alignment, and reinforcement using English-speaking annotators. There are exceptions to this rule. Chinese models like Qwen and DeepSeek outperform Western models when handling Chinese text and cultural context, and certain