In plain words: An AI's words and internal probes may not reflect how it actually computes, so security checks that read its self-reports can always be fooled. A safer sandbox instead tracks data flow, blocking model output from touching protected parts of a system.
Abstract
LLMs are trained to generate natural language. However, various strands of evidence indicate that an LLM's externalized linguistic outputs and mechanistically-extracted linguistic features can be an unreliable lens for understanding internal model computation. We introduce the term ``linguistic illegibility'' to broadly refer to scenarios in which an LLM's externalized or mechanistically-probed language artifacts fail to represent how the model actually thinks. We argue that the specter of linguistic illegibility is unavoidable for LLMs whose internal computations are not directly expressed via language, but rather math over activation spaces (with lossy translations between activation spaces and natural language happening at the bookends). If linguistic illegibility is always possible, then security mechanisms that rely on a model's linguistic self-reporting (e.g., chain-of-thought monitoring, constitutional self-critique, activation probing for linguistically-defined feature vectors) can never be completely sound; the model sandbox will always need isolation techniques whose guarantees do not depend on reading a model's linguistic state at all. We argue that observing a model's outputs using taint tracking is a promising approach for an effective sandbox: regardless of how a model linguistically self-reports, a taint tracking policy can define, a priori, various pieces of system state that should never be influenced by model-produced data. We also discuss several additional sandboxing mechanisms (e.g., robust virtualization, third-party auditing of sandboxing configurations) which collectively provide a critical floor beneath linguistic monitoring, and would have mitigated recent sandbox exploits by frontier models.
James Mickens
arXiv:2609.02852 · cs.LG, cs.CR · submitted Sep 2, 2026
abstract · pdf · html
I dislike this term because it doesn't explain where this "illegibility" is coming from. Models are post-trained towards non-linguistic goals with (mostly) non-linguistic rewards. A model's reasoning chain is reinforced if it leads to a correct answer or agentic goal. It doesn't need to be linguistically accurate and meanings can drift over training.