Anthropic has accused Chinese AI firms DeepSeek, Moonshot AI, and MiniMax of conducting large-scale "distillation attacks" to extract capabilities from its Claude models. The company alleges that over 24,000 fraudulent accounts were used to generate more than 16 million Claude exchanges, aiming to replicate model functionalities and potentially bypass safety measures. This accusation has sparked debate within the AI community, with some viewing it as a natural consequence of training on internet data, while others emphasize the unique risks posed by systematic output extraction, especially concerning tool use and safety control replication. AI
Summary written by None from 19 sources. How we write summaries →
IMPACT Raises concerns about intellectual property theft and safety bypass in frontier models, potentially impacting future model development and regulation.
RANK_REASON Major AI lab accuses competitors of large-scale data extraction and capability theft.