The story’s score and claims

Comments on No. 6 of 95 · arXiv · edition of Tuesday 10 November 2026

I Think, Therefore I Attack

Summary

Researchers at UT Austin's Applied Research Laboratories tested "thought injection," where untrusted text mimics a reasoning model's own chain of thought. In a containerized cyber range, payloads written in the target's reasoning style and wrapped in thinking tags reached 50% success on repository backdooring and 42% on reverse-shell exfiltration against MiniMax-M2.5. The same instructions phrased as ordinary prompt injection never succeeded. Effects of tags varied by model, with 50 runs per cell on two models. Tag stripping looks unreliable, and filters should scan reasoning output too.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules