LLMs can now identify public figures in images

A comparative analysis reveals significant differences in how major LLMs handle public figure identification in images. While ChatGPT and Claude refuse to identify people due to privacy safeguards, Google's Gemini, Meta's Llama, Mistral, and Qwen readily identify public figures with varying accuracy. The research demonstrates that OpenAI and Anthropic have stricter RLHF training around privacy, but these restrictions can be bypassed through specific prompt engineering techniques. Gemini shows the highest accuracy in identification tasks, likely due to Google's extensive image training data from search operations.