# The Invisible Editorial Layer: Formalizing Undisclosed Inference-Time Steering, Probability Placement, and the Attribution Problem in Deployed Language Models

arXiv: https://arxiv.org/abs/2608.24662 · HTML: https://mlsanity.com/paper/2608.24662/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 12 of its 33 references that are in the map.
- First posted on arXiv 25 Aug 2026.

## Builds on (most cited first, 12 of 12)

- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [Plug and Play Language Models: A Simple Approach to Controlled Text Generation](https://mlsanity.com/paper/1912.02164/) (2019)
- [A Watermark for Large Language Models](https://mlsanity.com/paper/2301.10226/) (2023)
- [DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Experts](https://mlsanity.com/paper/2105.03023/) (2021)
- [GeDi: Generative Discriminator Guided Sequence Generation](https://mlsanity.com/paper/2009.06367/) (2020)
- [FUDGE: Controlled Text Generation With Future Discriminators](https://mlsanity.com/paper/2104.05218/) (2021)
- [Black-Box Access is Insufficient for Rigorous AI Audits](https://mlsanity.com/paper/2401.14446/) (2024)
- [On the Conversational Persuasiveness of Large Language Models: A Randomized Controlled Trial](https://mlsanity.com/paper/2403.14380/) (2024)
- [From Transparency to Accountability and Back: A Discussion of Access and Evidence in AI Auditing](https://mlsanity.com/paper/2410.04772/) (2024)
- [Steering Language Models Before They Speak: Logit-Level Interventions](https://mlsanity.com/paper/2601.10960/) (2026)
- [Don't Change My View: Ideological Bias Auditing in Large Language Models](https://mlsanity.com/paper/2509.12652/) (2025)

## Abstract (arXiv)

Evaluations of generative language models frequently interpret observable behavioral traits, such as political stance, brand inclination, and normative framing, as manifestations of model weights, post-training alignment, or prompting. This interpretation risks conflating a foundation model with the multi-layered production system through which its outputs are ultimately served. Modern inference stacks support runtime interventions capable of modifying generation while model parameters remain frozen. We examine inference-time framing bias: systematic runtime steering of generated text toward institutional, ideological, or commercial frames without requiring changes to the underlying model parameters. We formalize the Inference Attribution Problem and establish an observational non-identifiability result showing that, under black-box observation alone, behaviorally equivalent deployed systems may arise from structurally distinct combinations of model parameters and inference policies. Consequently, observed behavioral bias does not uniquely identify the architectural layer responsible for it. We further characterize Probability Placement as a deployment pattern in which undisclosed commercial influence is embedded within an ostensibly organic assistant response through systematic probability-mass reallocation, distinguishing it from explicit token-auction mechanisms for generative advertising. Finally, we discuss implications for behavioral auditing, inference provenance, confidential computing, cryptographic attestation, the EU AI Act, the Digital Services Act, and advertising-disclosure principles. We argue that governance of generative systems must increasingly distinguish between auditing a model and auditing the deployed system that ultimately speaks.