Switch Languages, Bypass AI Safety: Hackers Exploit a Simple Trick

Hackers are evading artificial intelligence safety features by switching the language of their prompts. Instead of using English, they type commands in less common natural languages, such as Zulu or Welsh. AI makers design their systems to block harmful requests, like instructions for creating weapons or malware. However, many safety filters only work well in widely spoken languages. By using a rare language, hackers can trick the AI into ignoring its own rules. AI companies are now working to expand their safety checks to cover more languages. The challenge is large: there are thousands of languages, and building effective filters for each one takes time and resources. For now, the simple trick of switching languages remains a persistent loophole.