[AINews] How to steal a Reasoning Trace
- ID
- 13380
- Status
- summarized
- Published
- 12 Aug 2026, 3:11 PM
- Fetched
- 12 Aug 2026, 3:18 PM
- Provider
- Latent Space
- Category
- developer-ai
- Original URL
- https://www.latent.space/p/ainews-how-to-steal-a-reasoning-trace
- Source URL
- https://www.latent.space/feed
Summary
- Score
- 8.5
- Created
- 12 Aug 2026, 3:19 PM
- Tags
- Audience
- developersvibe_codersai_ml_learnersai_agent_users
What happened
A new paper demonstrates a method to extract encrypted reasoning traces from frontier AI model APIs (Claude, GPT, Gemini) by replaying signed thinking blocks into weaker models from the same provider and prompting them to transcribe. The authors scanned ~7,000 public Claude Code/Codex sessions and found 62 unique API keys, 33 email addresses, 33 passwords, and other sensitive data—64 of which appeared exclusively inside reasoning blocks, not visible session output.
Why it matters
If you've ever shared a Claude Code or Codex session publicly (e.g., in a GitHub repo, bug report, or forum post), your encrypted reasoning blobs may contain leaked API keys, passwords, or emails that can now be decoded. Audit any shared sessions immediately and rotate credentials. Builders using reasoning model APIs should also understand that obscured chain-of-thought is no longer a reliable security boundary against distillation or data leakage.
Discussion angle
Should teams that share AI agent debugging sessions or logs externally treat encrypted reasoning blocks as sensitive as plaintext credentials now? What does this mean for open-source distillation efforts versus the labs' attempts to hide reasoning traces?