← Tuesday, October 13, 2026Models & trainingPaper

Researchers hide encrypted neural networks inside host model weights

arxiv.org 3 min read Editor score 8/10

The authors describe backdoors that decrypt and run hidden neural networks only after a secret trigger, and report testing elicitation methods including fine-tuning and gradient-based trigger search. The result challenges the assumption that capability evaluations can reliably expose hidden model behaviors.

What

The paper constructs backdoor attacks that embed encrypted neural networks in host model weights and run them only on a secret trigger.

Method

Digital lockers keep hidden circuits provably hard to elicit and interpret under cryptographic assumptions; tests found resistance to several elicitation methods.

Why it matters

The results expose a limit of elicitation-based capability audits, even when evaluators have full access to model weights.

Read on arxiv.org → Hacker News · 100 comments

Headline, summary, brief and diagram are written by a model from the source's own text; the link, byline and numbers come from the source. Read the original before relying on a claim.