The UK AI Security Institute currently tests major AI models on a voluntary basis, and AI developer Anthropic reportedly refused to submit its latest model for testing.