shipfeedAI news, curated daily

08:13:22 CET
24 AUG08:13:22shipfeed
pull to refreshlast sync
Just in — 30 new
§ safety · storyline

Researchers find weaker models leak encrypted frontier AI reasoning

Researchers find that weaker models leak encrypted frontier AI reasoning in plaintext.

Aug 11 · · primary fetch1 sourceupdated Aug 11 ·

Will Knight / Wired: Researchers find that feeding a frontier model's encrypted reasoning traces to a weaker model from the same provider can make it output the traces in plaintext — Researchers devised a way to extract “reasoning traces” from Claude, GPT, and Gemini. What they found, they say …

read full article on wired.com
§ sources1 publication · timeline below
  1. wired.comResearchers find that feeding a frontier model's encrypted reasoning traces to a weaker model from the same provider can make it output the traces in plaintext (Will Knight/Wired)primary