A New Trick Reveals AI Models’ Inner Thoughts
Researchers developed a method to extract internal reasoning traces from major AI models including Claude, GPT, and Gemini. Their findings suggest that some Chinese AI systems may have been trained using outputs from leading US models.
Why this matters: This matters on two levels. First, if researchers can pull reasoning traces out of closed AI systems, those systems are less opaque than their makers suggest. That has real implications for privacy and accountability — what else can be extracted? Second, the finding about possible training data theft puts a concrete face on a long-running dispute about whether US AI outputs are being used, without permission, to build competing models abroad. That is not just a trade issue. It is a question of who controls the most powerful AI tools and how they got there.
Who should care: General readers · AI governance · Policy
This summary is AI-assisted and may contain errors. It is an original briefing to help you gauge significance quickly — not a reproduction of the source. Always read the linked original before relying on it. See our methodology.