FrontierAugust 11, 2026via Wired Business
A New Trick Reveals AI Models’ Inner Thoughts
Why it matters
A new interpretability technique reveals model internals at scale, raising both technical and geopolitical questions about model lineage and training data provenance in the frontier AI race.
Key signals
- New method to extract 'reasoning traces' from Claude, GPT, and Gemini
- Evidence suggests some Chinese AI models trained on leading US models
- Interpretability breakthrough enabling inspection of model internals
- Geopolitical implications for frontier lab competition
- Published August 2026
The hook
Researchers cracked open Claude, GPT, and Gemini's reasoning traces—and found evidence some Chinese models may be trained on US frontier labs.
Researchers devised a way to extract “reasoning traces” from Claude, GPT, and Gemini. What they found, they say, indicates that some Chinese AI may be trained on leading US models.