Researchers extract hidden reasoning from frontier AI models
Researchers devised a method to extract hidden reasoning traces from Claude, GPT, and Gemini via API, finding that Chinese model Kimi K3 produces strikingly similar outputs to Claude Opus 4.8 and GPT 5.6 Sol, suggesting possible distillation. The method can also recover personal info like passwords, a vulnerability now fixed.
Featured · Alexander Panfilov
2 sources
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Prompt search tool indexes 10,000+ image generation prompts
- Judge refuses to toss indie musician's lawsuit against Suno
- Nvidia sells first H200 chips in China, shipments below allowed total
- Model company acquisitions discussed
- Apple's Luce generates relightable 3D assets from single images