LLM Auditing

LLM: Large Language Model

Overview

Large language model auditing examines how models are trained and how they behave in use, looking for problems such as bias, privacy leakage, hallucinations, unsafe actions, and unreliable reasoning. Common approaches include controlled benchmarks and counterfactual tests, multi-turn red-teaming, membership-inference and provenance analysis, human review, and probes of model activations or weights. Audits increasingly examine whole systems—including prompts, tools, interfaces, and model updates—rather than judging only a model’s final answer. A central challenge is that results can shift with prompts, access methods, versions, and evaluators, while behavioral scores or internal signals alone may not establish why a failure occurred or whether it will recur in deployment.

Latest papers

723 papers79 papers in the last four weeks, down 15% on the four weeks before. 0.6% of all new papers.

May 21, 2025

Auditing Prompt Injection Defenses Before They Reach Users: What Strong Adversaries Reveal

Large language models are increasingly deployed in user-facing products that handle sensitive personal data while interacting with untrusted external content. This makes them a prime target for prompt injection attacks, in which an adversary injects malicious instructions in that content to hijack the model. Fine-tuning-based defenses have been proposed to mitigate these attacks by training models to follow benign user instructions and ignore injected ones. Lacking formal guarantees, these defenses are evaluated empirically against the most effective realistic attacks available at the time, often reporting near-zero attack success rates -- giving developers, and the users who depend on them, a false sense of security and leaving open the possibility that future, more capable attacks could circumvent them and cause severe consequences in widely deployed systems. In this paper, we argue that developers should stress-test their defenses before release, and we introduce an auditing framework that does so without requiring defenders to invent the next state-of-the-art attack. Our key insight is that the fine-tuning process itself yields useful artifacts: intermediate checkpoints that serve as stepping stones for optimization-based attacks. Although these artifacts are unavailable to real attackers, we show that the vulnerabilities they expose are genuine and can be exploited by realistic adversaries. Instantiating our framework with optimization-based attacks such as GCG and BEAST, we raise attack success rates against state-of-the-art defenses from near zero to as high as 96%, revealing weaknesses missed by prior evaluations. Our results show that current evaluation practices can substantially overestimate robustness, and that auditing defenses under strong adversaries before deployment shortens the feedback loop and helps prevent attacks from causing real-world harm to users.
May 6, 2025

Say It Another Way: Auditing LLMs with a User-Grounded Automated Paraphrasing Framework

Large language models (LLMs) are highly sensitive to subtle changes in prompt phrasing, posing challenges for reliable auditing. Prior methods often apply unconstrained prompt paraphrasing, which risk missing linguistic and demographic factors that shape authentic user interactions. We introduce AUGMENT (Automated User-Grounded Modeling and Evaluation of Natural Language Transformations), a framework for generating controlled paraphrases, grounded in user behaviors. AUGMENT leverages linguistically informed rules and enforces quality through checks on instruction adherence, semantic similarity, and realism, ensuring paraphrases are both reliable and meaningful for auditing. Through case studies on the BBQ and MMLU datasets, we show that controlled paraphrases uncover systematic weaknesses that remain obscured under unconstrained variation. These results highlight the value of the AUGMENT framework for reliable auditing.
Apr 24, 2025

Beyond Public Access in LLM Pre-Training Data

Using a legally obtained dataset of 34 copyrighted O'Reilly Media books (13,962 paragraphs), we apply the DE-COP membership inference attack to test whether OpenAI models show recognition of non-public (pay-walled) content they were not licensed to train on. Our main contribution is examining non-public books, which allows us to consider potential access violations and not just training inclusion. Our results suggest that GPT-4o, OpenAI's more recent and capable model in our sample, exhibits patterns consistent with recognition of non-public books, with an AUROC score of 0.82 (95% bootstrapped CI: 0.60-0.96) -- where above 50% is better than random chance. GPT-3.5 Turbo (trained two years prior) and the smaller GPT-4o Mini both score near chance recognition on non-public data (0.54 and 0.56, respectively). Results are robust to independent paraphrasers (Gemma and Qwen) and leave-one-book-out sensitivity analyses. We also test both GPT-4o and GPT-4o mini, with the same October 2023 training cutoff date, to ensure the model is not simply identifying language shifts over time. These findings support disclosure requirements for pre-training data provenance and formal licensing markets for training content. Results remain preliminary given the small book sample, possible contamination of post-cutoff controls, and differences across models in size, architecture, and training composition.