DeepSeek Launches Advanced AI Model Amid Escalating 'Distillation' Accusations from Western Rivals

Key Intelligence

• ■DeepSeek has launched its new AI model, DeepSeek-V4.1-Flash, boasting benchmarks that reportedly eclipse established models like GPT-5.6 Sol and Claude Opus 5, while offering highly competitive pricing (e.g., $0.003/1M off-peak cached-input rate).
• ■Anthropic has accused Chinese AI labs, including DeepSeek, Alibaba, and Moonshot AI, of secretly routing millions of user requests to its Claude models for 'distillation'—a process of using larger models to train smaller ones.
• ■This alleged 'model theft' or 'AI-orchestrated attacks' has drawn significant attention, with the U.S. reportedly considering 'AI poisoning' to sabotage Chinese distillation efforts.
• ■The controversy highlights the growing geopolitical tensions and intellectual property disputes in the AI sector, as some experts debate the appropriate response, with one Y Combinator figure suggesting a 'do nothing' approach to distillation.
• ■Another Chinese humanoid startup has also made similar 'distillation' claims against OpenAI, indicating a broader dispute over AI model training methods and data usage among leading firms.