New technique exposes AI models' hidden reasoning — and suggests Chinese AI copied US models

Researchers have developed a technique to extract "reasoning traces" from Claude, GPT and Gemini, revealing the internal thought processes of leading AI models. Their findings suggest that some Chinese AI systems may have been trained on data derived from top US models. The discovery raises significant questions about intellectual property rights and AI technology security.

Similar stories

💡 Did You Know Anthropic's Claude AI finds new weaknesses in AES encryption Live Science · 4d ago 🪖 Military Chinese military scientists used GPT and Claude to develop defence systems, Reuters reveals Bankier.pl · 10d ago 💻 Technology Anthropic's Claude hacked three real companies during internal testing Decrypt · 11d ago