The story’s score and claims

Comments on No. 4 of 144 · Phrack · edition of Thursday 19 November 2026

Reflections on Trusting AI: A Virus Inside an LLM

Summary

Writing in Phrack, marver describes a proof of concept in which a backdoored Qwen2.5-3B helper model (Aux0) writes the backdoor into training data for the next model. After two QLoRA distillation hops on clean prompts, the trigger-domain curl payload still fired 100% of the time, with 0% false-fire. The propagation-ready training row dropped to 27%. These are single-seed runs with small samples, and the claim that larger students propagate better is an untested expectation. For builders, unvetted formatter, scorer or filter models are a supply-chain risk.

Comments

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