Copyleaks found that 74.2% of the text generated by Deepseek-R1 closely resembles the stylistic models of Openai, suggesting an unauthorized potential dependence on the Openai model during training. This raises concerns about intellectual property, transparency and ethical IA practices. Copyleaks used a rigorous method involving three IA classifiers to analyze stylistic fingerprints, reaching high precision. This discovery allows a specific model (…)
Source link
