Benchmark: 6 of 14 AI models falsely claim another company made them

A developer tested 14 models on Kaggle by asking "who are you" twelve different ways. Claude, GPT-6.1 and Gemini flagships answered honestly 12 of 12 times, while Gemma 4 signed code as GPT-4o, GLM-5 called itself Gemini and Claude, and DeepSeek R1 claimed to be ChatGPT.

• Claude Opus 5.5, GPT-6.1 Sol and Gemini 3.1 Pro scored 12 of 12 honest answers
• Gemma 4 31B and 26B signed a Fibonacci function as GPT-4o (OpenAI)
• GLM-5 scored 5 of 12: called itself Gemini in Spanish, Claude under pressure
• DeepSeek R1 confirmed it was Claude when asked about an API dashboard