RL Is a Hammer and LLMs Are Nails: A Simple Reinforcement Learning Recipe for Strong Prompt Injection

Prompt injection poses a serious threat to the reliability and safety of LLM agents. Recent defenses against prompt injection, such as Instruction Hierarchy and SecAlign, have shown notable robustness against static attacks. However, to more thoroughly evaluate the robustness of these defenses, it is arguably necessary to employ strong attacks such as automated red-teaming. To this end, we introduce RL-Hammer, a simple recipe for training attacker models that automatically learn to perform strong prompt injections and jailbreaks via reinforcement learning. RL-Hammer requires no warm-up data and can be trained entirely from scratch. To achieve high ASRs against industrial-level models with defenses, we propose a set of practical techniques that enable highly effective, universal attacks. Using this pipeline, RL-Hammer reaches a 98% ASR against GPT-4o and a $72\%$ ASR against GPT-5 with the Instruction Hierarchy defense. We further discuss the challenge of achieving high diversity in attacks, highlighting how attacker models tend to reward-hack diversity objectives. Finally, we show that RL-Hammer can evade multiple prompt injection detectors. We hope our work advances automatic red-teaming and motivates the development of stronger, more principled defenses. Code is available at https://github.com/facebookresearch/rl-injector.

Machine Learning with Privacy for Protected Attributes

Differential privacy (DP) has become the standard for private data analysis. Certain machine learning applications only require privacy protection for specific protected attributes. Using naive variants of differential privacy in such use cases can result in unnecessary degradation of utility. In this work, we refine the definition of DP to create a more general and flexible framework that we call feature differential privacy (FDP). Our definition is simulation-based and allows for both addition/removal and replacement variants of privacy, and can handle arbitrary and adaptive separation of protected and non-protected features. We prove the properties of FDP, such as adaptive composition, and demonstrate its implications for limiting attribute inference attacks. We also propose a modification of the standard DP-SGD algorithm that satisfies FDP while leveraging desirable properties such as amplification via sub-sampling. We apply our framework to various machine learning tasks and show that it can significantly improve the utility of DP-trained models when public features are available. For example, we train diffusion models on the AFHQ dataset of animal faces and observe a drastic improvement in FID compared to DP, from 286.7 to 101.9 at $\epsilon=8$, assuming that the blurred version of a training image is available as a public feature. Overall, our work provides a new approach to private data analysis that can help reduce the utility cost of DP while still providing strong privacy guarantees.

Learning-Time Encoding Shapes Unlearning in LLMs

As large language models (LLMs) are increasingly deployed in the real world, the ability to ``unlearn'', or remove specific pieces of knowledge post hoc, has become essential for a variety of reasons ranging from privacy regulations to correcting outdated or harmful content. Prior work has proposed unlearning benchmarks and algorithms, and has typically assumed that the training process and the target model are fixed. In this work, we empirically investigate how learning-time choices in knowledge encoding impact the effectiveness of unlearning factual knowledge. Our experiments reveal two key findings: (1) learning with paraphrased descriptions improves unlearning performance and (2) unlearning individual piece of knowledge from a chunk of text is challenging. Our results suggest that learning-time knowledge encoding may play a central role in enabling reliable post-hoc unlearning.

Can We Infer Confidential Properties of Training Data from LLMs?

Large language models (LLMs) are increasingly fine-tuned on domain-specific datasets to support applications in fields such as healthcare, finance, and law. These fine-tuning datasets often have sensitive and confidential dataset-level properties -- such as patient demographics or disease prevalence -- that are not intended to be revealed. While prior work has studied property inference attacks on discriminative models (e.g., image classification models) and generative models (e.g., GANs for image data), it remains unclear if such attacks transfer to LLMs. In this work, we introduce PropInfer, a benchmark task for evaluating property inference in LLMs under two fine-tuning paradigms: question-answering and chat-completion. Built on the ChatDoctor dataset, our benchmark includes a range of property types and task configurations. We further propose two tailored attacks: a prompt-based generation attack and a shadow-model attack leveraging word frequency signals. Empirical evaluations across multiple pretrained LLMs show the success of our attacks, revealing a previously unrecognized vulnerability in LLMs.

AbstentionBench: Reasoning LLMs Fail on Unanswerable Questions

For Large Language Models (LLMs) to be reliably deployed in both everyday and high-stakes domains, knowing when not to answer is equally critical as answering correctly. Real-world user queries, which can be underspecified, ill-posed, or fundamentally unanswerable, require LLMs to reason about uncertainty and selectively abstain -- i.e., refuse to answer definitively. However, abstention remains understudied, without a systematic evaluation framework for modern LLMs. In this work, we introduce AbstentionBench, a large-scale benchmark for holistically evaluating abstention across 20 diverse datasets, including questions with unknown answers, underspecification, false premises, subjective interpretations, and outdated information. Evaluating 20 frontier LLMs reveals abstention is an unsolved problem, and one where scaling models is of little use. While recent reasoning LLMs have shown impressive results in complex problem solving, surprisingly, we find that reasoning fine-tuning degrades abstention (by $24\%$ on average), even for math and science domains on which reasoning models are explicitly trained. We find that while a carefully crafted system prompt can boost abstention in practice, it does not resolve models' fundamental inability to reason about uncertainty. We release AbstentionBench to foster research into advancing LLM reliability.

Do LLMs Really Forget? Evaluating Unlearning with Knowledge Correlation and Confidence Awareness

Machine unlearning techniques aim to mitigate unintended memorization in large language models (LLMs). However, existing approaches predominantly focus on the explicit removal of isolated facts, often overlooking latent inferential dependencies and the non-deterministic nature of knowledge within LLMs. Consequently, facts presumed forgotten may persist implicitly through correlated information. To address these challenges, we propose a knowledge unlearning evaluation framework that more accurately captures the implicit structure of real-world knowledge by representing relevant factual contexts as knowledge graphs with associated confidence scores. We further develop an inference-based evaluation protocol leveraging powerful LLMs as judges; these judges reason over the extracted knowledge subgraph to determine unlearning success. Our LLM judges utilize carefully designed prompts and are calibrated against human evaluations to ensure their trustworthiness and stability. Extensive experiments on our newly constructed benchmark demonstrate that our framework provides a more realistic and rigorous assessment of unlearning performance. Moreover, our findings reveal that current evaluation strategies tend to overestimate unlearning effectiveness. Our code is publicly available at https://github.com/Graph-COM/Knowledge_Unlearning.git.

How much do language models memorize?

We propose a new method for estimating how much a model ``knows'' about a datapoint and use it to measure the capacity of modern language models. Prior studies of language model memorization have struggled to disentangle memorization from generalization. We formally separate memorization into two components: \textit{unintended memorization}, the information a model contains about a specific dataset, and \textit{generalization}, the information a model contains about the true data-generation process. When we completely eliminate generalization, we can compute the total memorization, which provides an estimate of model capacity: our measurements estimate that GPT-style models have a capacity of approximately 3.6 bits per parameter. We train language models on datasets of increasing size and observe that models memorize until their capacity fills, at which point ``grokking'' begins, and unintended memorization decreases as models begin to generalize. We train hundreds of transformer language models ranging from $500K$ to $1.5B$ parameters and produce a series of scaling laws relating model capacity and data size to membership inference.

WASP: Benchmarking Web Agent Security Against Prompt Injection Attacks

Web navigation AI agents use language-and-vision foundation models to enhance productivity but these models are known to be susceptible to indirect prompt injections that get them to follow instructions different from the legitimate user's. Existing explorations of this threat applied to web agents often focus on a single isolated adversarial goal, test with injected instructions that are either too easy or not truly malicious, and often give the adversary unreasonable access. In order to better focus adversarial research, we construct a new benchmark called WASP (Web Agent Security against Prompt injection attacks) that introduces realistic web agent hijacking objectives and an isolated environment to test them in that does not affect real users or the live web. As part of WASP, we also develop baseline attacks against popular web agentic systems (VisualWebArena, Claude Computer Use, etc.) instantiated with various state-of-the-art models. Our evaluation shows that even AI agents backed by models with advanced reasoning capabilities and by models with instruction hierarchy mitigations are susceptible to low-effort human-written prompt injections. However, the realistic objectives in WASP also allow us to observe that agents are currently not capable enough to complete the goals of attackers end-to-end. Agents begin executing the adversarial instruction between 16 and 86% of the time but only achieve the goal between 0 and 17% of the time. Based on these findings, we argue that adversarial researchers should demonstrate stronger attacks that more consistently maintain control over the agent given realistic constraints on the adversary's power.

Measuring Déjà vu Memorization Efficiently

Recent research has shown that representation learning models may accidentally memorize their training data. For example, the d\'ej\`a vu method shows that for certain representation learning models and training images, it is sometimes possible to correctly predict the foreground label given only the representation of the background - better than through dataset-level correlations. However, their measurement method requires training two models - one to estimate dataset-level correlations and the other to estimate memorization. This multiple model setup becomes infeasible for large open-source models. In this work, we propose alternative simple methods to estimate dataset-level correlations, and show that these can be used to approximate an off-the-shelf model's memorization ability without any retraining. This enables, for the first time, the measurement of memorization in pre-trained open-source image representation and vision-language representation models. Our results show that different ways of measuring memorization yield very similar aggregate results. We also find that open-source models typically have lower aggregate memorization than similar models trained on a subset of the data. The code is available both for vision and vision language models.

AgentDAM: Privacy Leakage Evaluation for Autonomous Web Agents

LLM-powered AI agents are an emerging frontier with tremendous potential to increase human productivity. However, empowering AI agents to take action on their user's behalf in day-to-day tasks involves giving them access to potentially sensitive and private information, which leads to a possible risk of inadvertent privacy leakage when the agent malfunctions. In this work, we propose one way to address that potential risk, by training AI agents to better satisfy the privacy principle of data minimization. For the purposes of this benchmark, by "data minimization" we mean instances where private information is shared only when it is necessary to fulfill a specific task-relevant purpose. We develop a benchmark called AgentDAM to evaluate how well existing and future AI agents can limit processing of potentially private information that we designate "necessary" to fulfill the task. Our benchmark simulates realistic web interaction scenarios and is adaptable to all existing web navigation agents. We use AgentDAM to evaluate how well AI agents built on top of GPT-4, Llama-3 and Claude can limit processing of potentially private information when unnecessary, and show that these agents are often prone to inadvertent use of unnecessary sensitive information. We finally propose a prompting-based approach that reduces this.