Elon Musk criticises Anthropic over allegations about AI model training practices

Elon Musk has launched a sharp critique against Anthropic, following the company's allegations that several Chinese AI firms, including DeepSeek, have been illicitly obtaining data from its Claude AI model. This public dispute unfolded on X (formerly Twitter), where Musk accused the Dario Amodei-led organization of engaging in unethical practices. In a detailed blog post, Anthropic claimed that three Chinese AI companies, namely DeepSeek, Moonshot AI, and MiniMax, have been training their models using Claude's responses. The accusations suggest that these firms created over 24,000 fake accounts, resulting in more than 16 million interactions with Claude to enhance their own AI systems. Following these assertions, Musk took to X to declare, "Anthropic is guilty of stealing training data at a massive scale and has had to pay multi-billion dollar settlements for their theft. This is just a fact." He further referenced X Community Notes that indicated Anthropic previously settled a $1.5 billion lawsuit related to the alleged unauthorized use of data to train Claude, as well as facing a potential $3 billion lawsuit for illegally downloading over 20,000 songs. The Community Notes highlighted that the firm allegedly built its models on stolen creative works and questioned the company's integrity when it comes to its own data usage. Musk's comments were not only a defense of his position, but they also served to remind Anthropic to reflect on its data practices before casting stones at others. The issue of distillation attacks, where less capable models are developed based on the outputs of more advanced systems, has emerged as a significant concern in the AI sector. Major tech players like OpenAI and Google have also reported similar incidents of potential data misuse. In response, these companies are taking steps to bolster defenses against unauthorized access to their AI models.