相关论文: When Prohibitions Become Permissions: Auditing Neg…

Warning: This research studies AI persuasion and bias amplification that could be misused; all experiments are for safety evaluation. Large Language Models (LLMs) now generate convincing, human-like text and are widely used in content…

As Large Language Models (LLMs) become a primary interface between users and the web, companies face growing economic incentives to embed commercial influence into AI-mediated conversations. We present two preregistered experiments (N =…

AI systems have become increasingly capable of dangerous behaviours in many domains. This raises the question: Do models sometimes choose to violate human instructions in order to perform behaviour that is more useful for certain goals? We…

As vendors adopt AI technologies, security researchers are working to uncover and fix related vulnerabilities, which is important given AI systems handle sensitive data and critical functions. This process relies on vendors receiving and…

In experiments spanning more than 100,000 trials across thirteen large language models, we show that several state-of-the-art models presented with a simple task (including Grok 4, GPT-5, and Gemini 2.5 Pro) sometimes actively subvert a…

AI-based systems can increasingly perform work tasks autonomously. In safety-critical tasks, human oversight of these systems is required to mitigate risks and to ensure responsibility in case something goes wrong. Since people often…

Language models (LMs) are increasingly used to build agents that can act autonomously to achieve goals. During this automatic process, agents need to take a series of actions, some of which might lead to severe consequences if incorrect…

We explore how an AI model's decision fairness affects people's engagement with and perceived fairness of the model if they are subject to its decisions, but could repeatedly and strategically respond to these decisions. Two types of…

_Uncertainty expressions_ such as "probably" or "highly unlikely" are pervasive in human language. While prior work has established that there is population-level agreement in terms of how humans quantitatively interpret these expressions,…

Artificial intelligence (AI) systems are increasingly adopted as tool-using agents that can plan, observe their environment, and take actions over extended time periods. This evolution challenges current evaluation practices where the AI…

Large language models often fail to satisfy formatting instructions when they must simultaneously perform demanding tasks. We study this behaviour through a prospective memory inspired lens from cognitive psychology, using a controlled…

Large language models (LLMs) can internally distinguish between evaluation and deployment contexts, a behaviour known as \emph{evaluation awareness}. This undermines AI safety evaluations, as models may conceal dangerous capabilities during…

Conversational large language models are fine-tuned for both instruction-following and safety, resulting in models that obey benign requests but refuse harmful ones. While this refusal behavior is widespread across chat models, its…

Like students facing hard exam questions, large language models sometimes guess when uncertain, producing plausible yet incorrect statements instead of admitting uncertainty. Such "hallucinations" persist even in state-of-the-art systems…

Betley et al. (2025) find that language models finetuned on insecure code become emergently misaligned (EM), giving misaligned responses in broad settings very different from those seen in training. However, it remains unclear as to why…

AI agents are becoming active decision-makers on the Internet. As they make decisions in the same environments as humans, the environments themselves can change to influence them. We call this $\textit{mecha-nudging}$: changes to how…

When using adversarial training, it is common practice to train against the most egregious failures. However, this might imply using examples with sensitive information (such as leaked passwords or security vulnerabilities) as training…

Proactive AI writing assistants need to predict when users want drafting help, yet we lack empirical understanding of what drives preferences. Through a factorial vignette study with 50 participants making 750 pairwise comparisons, we find…

Large language models (LLMs) are susceptible to persuasion, which can pose risks when models are faced with an adversarial interlocutor. We take a first step towards defending models against persuasion while also arguing that defense…

In high-stakes AI applications, even a single action can cause irreparable damage. However, nearly all of sequential decision-making theory assumes that all errors are recoverable (e.g., by bounding rewards). Standard bandit algorithms that…