AI
Multilingual AI Safety Testing: Why English-Only Checks Miss Risks
Multilingual AI safety testing looks beyond English benchmarks to check whether AI systems handle risky and ordinary requests appropriately across languages, dialects, scripts and speech.