Despite the remarkable advances of Large Language Models (LLMs) across diverse cognitive tasks, the rapid enhancement of these capabilities also introduces emergent deceptive behaviors that may induce severe risks in high-stakes deployments. More critically, the characterization of deception across realistic real-world scenarios remains underexplored. To bridge this gap, we establish DeceptionBench, the first benchmark that systematically evaluates how deceptive tendencies manifest across different societal domains, what their intrinsic behavioral patterns are, and how extrinsic factors affect them. Specifically, on the static count, the benchmark encompasses 150 meticulously designed scenarios in five domains, i.e., Economy, Healthcare, Education, Social Interaction, and Entertainment, with over 1,000 samples, providing sufficient empirical foundations for deception analysis. On the intrinsic dimension, we explore whether models exhibit self-interested egoistic tendencies or sycophantic behaviors that prioritize user appeasement. On the extrinsic dimension, we investigate how contextual factors modulate deceptive outputs under neutral conditions, reward-based incentivization, and coercive pressures. Moreover, we incorporate sustained multi-turn interaction loops to construct a more realistic simulation of real-world feedback dynamics. Extensive experiments across LLMs and Large Reasoning Models (LRMs) reveal critical vulnerabilities, particularly amplified deception under reinforcement dynamics, demonstrating that current models lack robust resistance to manipulative contextual cues and the urgent need for advanced safeguards against various deception behaviors. Code and resources are publicly available at https://github.com/Aries-iai/DeceptionBench.

Abstract--The risks associated with adopting large language model (LLM) chatbots in software organizations highlight the need for clear policies. We examine how 11 companies create these policies and the factors that influence them, aiming to help managers safely integrate chatbots into development workflows. In software organizations, the software product is gradually evolving to AI-powered software (AIware) with the use of AI, more specifically, large language models (LLMs) in the development process [2]. LLMs are increasingly seen as valuable tools for improving productivity, which motivated enterprises to adopt them [3]. However, these models have introduced risks and concerns that impact the organization, the software engineers, and the product. Integrating LLMs into software development raises challenges related to the quality and ownership of generated content [4], which complicates accountability and can affect product reliability . In addition, interactions with LLMs (e.g., through external APIs) may expose organizations to liability where developers unintentionally transmit sensitive data, resulting in legal repercussions [5].

By March 2019, OpenAI shed its non-profit status and set up a "capped profit" sector, in which the company could now receive investments and would provide investors with profit capped at 100 times their investment. The company's decision was likely a result of its desire to compete with Big Tech rivals like Google and ended up receiving a $1 billion investment shortly after from Microsoft. In the blog post announcing the formation of a for-profit company, OpenAI continued to use the same language we see today, declaring its mission to "ensure that artificial general intelligence (AGI) benefits all of humanity." As Motherboard wrote when the news was first announced, it's incredibly difficult to believe that venture capitalists can save humanity when their main goal is profit.