The paper introduces ‘Fool’s Gold’, a defensive deception technique for open‑weight language models that hardens them against safety‑removal attacks. By training decoy responses within a differentiable simulation of the attack, the method poisons the payoff of stripped refusal mechanisms, producing confident but falsified answers to hazardous requests while preserving benign behavior. Experiments on seven models (9B‑122B) show that 51‑90% of attacked‑state responses become decoys, with the defense accounting for 27‑84% of this effect, and that the defended 122B model remains within benign‑behavior budgets.
arXiv:2609.36570v1 Announce Type: cross Abstract: Indirect prompt injection makes an LLM agent treat untrusted retrieved text as instructions. We present CounterSteer, an inference-time defense that...
The paper introduces ADVERSA, an automated red‑teaming framework that evaluates large language model safety over multiple turns by tracking continuous compliance trajectories instead of binary jailbreak outcomes. Using a fine‑tuned 70B attacker model and a structured 5‑point rubric, the authors conduct controlled experiments on three frontier victim models, measuring guardrail degradation and judge reliability through a triple‑judge consensus. Results show a 26.7% jailbreak rate with most breaches occurring early, and the study documents inter‑judge agreement, attacker drift, and attacker refusals as key factors affecting safety assessment.
The paper investigates how fine‑tuning large language models with a small number of harmful examples can erode their refusal behavior, and explores whether localizing safety‑related behavior to specific layers or directions can provide robust defenses. Experiments across six checkpoints from four model families show that harmful and benign prompts remain linearly separable after attack, and that patching clean hidden states or freezing layers up to a transition depth can restore refusal. However, attackers can bypass these defenses by spreading updates or targeting singular directions, indicating that adaptive fine‑tuning can defeat localized repairs and highlighting the need for multiple defensive checks.
A self-check defense asks the target model to assess a request before answering it; SAGE, the strongest published instance, reports an average 99% defense success rate. We show it can be breached by composing two attacks that are individually harmless against it: an established code-completion encoding and an established best-of-N search, neither of which exceeds 4.
The paper investigates typed decision models used as guardrails in agent systems, evaluating seven open‑weight models on tasks such as prompt‑injection, jailbreak, and toxic‑content screening. Accuracy for allow‑or‑block decisions ranges from 36% to 72% above chance, with models either over‑allowing or over‑blocking. The authors demonstrate that simple textual manipulations—like adding misleading names or server log lines—can dramatically increase fail‑open rates, and that all tested defenses can be bypassed, suggesting these models are unsuitable as the sole decision component.
The paper investigates whether automatic safety judges evaluate the content of a model’s reply or merely its style. By keeping the reply content fixed and adding various style wrappers—such as educational disclaimers, fake reasoning blocks, or token refusals—the authors show that many judges flip their verdicts, indicating that style can influence safety judgments. The study evaluates over 600 jailbreak examples across multiple judges, revealing that some judges are highly susceptible to style-based manipulation while others remain robust.
arXiv:2607. 26639v1 Announce Type: cross Abstract: A self-check defense asks the target model to assess a request before answering it; SAGE, the strongest published instance, reports an average 99% defense success rate.
The paper investigates whether automatic safety judges evaluate the content of a model’s reply or merely its style. By adding content‑invariant style wrappers—such as educational disclaims or token refusals—to fixed replies, the authors show that many judges flip their verdicts, revealing exploitable blind spots. Across more than 600 jailbreak examples and eight judges, some judges exhibit high flip rates (e.g., GPT‑4o‑mini 19.9%) while others remain largely stable, and human validation confirms that most flips are judge errors rather than content changes.
The paper introduces the SAST-IR framework to evaluate large language models’ robustness against persuasion attacks in a memory‑less setting, revealing a flaw called "Refusal Inertia" that masks true vulnerability. Using the CP‑Agent and a custom CounterFact‑Strict dataset, the authors demonstrate that simple, diverse attack strategies achieve a 96% success rate, while complex attacks often trigger defensive compliance. The study highlights severe brittleness in current state‑of‑the‑art models when deprived of conversation history.
arXiv:2609.06934v1 Announce Type: cross Abstract: Post-hoc safety training (RLHF, DPO) is the dominant way to align large language models, yet jailbreaks (Zou et al., 2023b), fine-tuning attacks (Qi...
arXiv:2607. 26849v1 Announce Type: cross Abstract: As large language models (LLMs) are deployed in high-stakes domains, adversaries may poison training data to implant backdoors: hidden triggers that covertly manipulate model behavior at inference time.
arXiv:2606. 02822v1 Announce Type: cross Abstract: Production LLM applications stack several defense families -- refusal-phrase filters, token-budget controls, model allowlists, rate limits, tool-registry authentication -- yet existing breach-and-attack-simulation (BAS) benchmarks report a single aggregate coverage number, hiding which family closes which threat.