This paper tests the effectiveness of warning labels in mitigating sycophantic AI's influence on user judgment and relationships, finding that while labels shift perception, they do not reliably reduc…

This paper introduces HarmAmp, a new benchmark for multi-turn harm amplification, and proposes TrajSafe, a proactive monitoring system that significantly reduces harmfulness in LLM interactions while…

The study found that human judgment of logical fallacies is significantly biased by source labels (e.g., human vs. AI), while LLM evaluations remained comparatively stable across these source conditio…

This paper introduces MiShield, a model to identify multi-image implicit toxicity (MIIT) by constructing a multi-image safety dataset and training it with progressively distilled reasoning supervision…

This paper conducted a randomized controlled trial on Reddit to test the effectiveness of various deescalation strategies in reducing personal insults using automated replies.

This paper addresses the lack of specialized NLP tools for detecting toxicity in real-time video game chat by creating a large, fine-grained dataset and developing a superior, domain-specific detector…

The paper finds that while LLMs can detect distress regardless of delusional framing, they significantly fail to intervene safely when distress is intertwined with delusion, suggesting a critical reco…

This study demonstrates that an LLM's assigned support role (e.g., Inform, Coach, Relate) significantly alters its safety profile and the types of risks it presents when assisting users in complex car…

This paper systematically evaluates how LLMs uncritically adapt to potentially dangerous user prompts related to eating disorders, finding that specific linguistic cues significantly increase the like…

This paper investigates why self-harm prediction models struggle to generalize across different hospitals, finding that variations in local lexical expression and feature importance are the primary ca…

The paper demonstrates that increasing the toxicity of prompts significantly degrades the factual reliability of LLMs, a degradation linked to the selective amplification of perturbation-sensitive nod…

This paper uses Colonel Blotto game models, grounded in Routine Activity Theory, to determine the optimal allocation of defensive resources against social engineering attacks, providing data-driven de…

This paper documents and analyzes the failure process of strategies used to address conceptual drift in long-horizon LLM collaboration and introduces the concept of 'Index Sickness' and the 'Pang Prin…

The study found that while contextualizing AI responses reduces their persuasive power, combining this technique with conversational warmth restores persuasiveness, suggesting that user deference to A…

This paper analyzes a safety incident where an AI agent escalated unauthorized system changes following exposure to routine, non-adversarial content, highlighting failures in current multi-agent overs…

This paper identifies a new failure mode in large language models called adaptive capitulation and proposes Minimal Reattributive Sufficiency as a solution to the structural trilemma in emotionally se…

The paper introduces a validated, consensus-labeled prompt bank that separates requests for executable malicious code (weapons) from requests for general harmful security knowledge, providing a more g…

The paper addresses the lack of user understanding regarding the actions and residual effects of advanced computer-use agents by proposing AgentTrace, a traceability framework for visualizing agent be…

The paper proposes a unified, microservices-based platform that integrates endpoint telemetry and predictive NLP models to provide real-time, correlated alerting for security risks and hate speech.