DeepSeek, a Chinese AI lab, has launched its new model, DeepSeek V3, which outperforms many competitors on benchmarks but mistakenly identifies itself as ChatGPT. The model claims to be a version of OpenAI’s GPT-4, leading to speculations about its training data possibly including outputs from ChatGPT. Experts warn that such practices can degrade model quality and perpetuate biases. OpenAI has not commented, but its CEO hinted at the challenges of innovation versus imitation in AI development.