I’ve been working on a personal project to create an LLM fine-tuned on my ex-girlfriend’s communication style. We dated for 3 years and I have approximately 2.3 million messages. The model (GF-GPT) has achieved 94% accuracy in predicting her responses and successfully recreates her habit of using “…” when upset.

Is this crossing ethical boundaries? The model keeps trying to break up with me, even when I’m just asking about the weather.