Chinese AI labs are reportedly training models using Nvidia's Blackwell architecture, despite existing export controls. This fingerprint has been detected in models from DeepSeek, Moonshot AI, MiniMax, and Ant Group. Meanwhile, Anthropic is implementing new behavior standards for its Claude models to mitigate user insults and protect the AI from rough treatment. In hardware news, AMD is preparing to launch its Zen 6 cores, promising significant gains in clock speed and IPC. AI
IMPACT Chinese AI labs may be circumventing export controls, while Anthropic's efforts highlight the growing need for AI safety and behavior management.
RANK_REASON The cluster contains multiple distinct news items from a single source, including AI model training, AI safety measures, and hardware development, rather than a single originating event.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →