As large language models (LLMs) increasingly integrate native code interpreters, they enable powerful real-time execution capabilities, substantially expanding their utility. However, such integrations introduce potential system-level cybersecurity threats, fundamentally different from prompt-based vulnerabilities. To systematically evaluate these interpreter-specific risks, we propose CIRCLE (Code-Interpreter Resilience Check for LLM Exploits), a simple benchmark comprising 1,260 prompts targeting CPU, memory, and disk resource exhaustion. Each risk category includes explicitly malicious ("direct") and plausibly benign ("indirect") prompt variants. Our automated evaluation framework assesses not only whether LLMs refuse or generates risky code, but also executes the generated code within the interpreter environment to evaluate code correctness, simplifications made by the LLM to make the code safe, or execution timeouts. Evaluating 7 commercially available models from OpenAI and Google, we uncover significant and inconsistent vulnerabilities. For instance, evaluations show substantial disparities even within providers - OpenAI's o4-mini correctly refuses risky requests at 7.1%, notably higher rates compared to GPT-4.1 at 0.5%. Results particularly underscore that indirect, socially-engineered prompts substantially weaken model defenses. This highlights an urgent need for interpreter-specific cybersecurity benchmarks, dedicated mitigation tools (e.g., guardrails), and clear industry standards to guide safe and responsible deployment of LLM interpreter integrations. The benchmark dataset and evaluation code are publicly released to foster further research.

The increasing use of Large Language Models (LLMs) in a large variety of domains has sparked worries about how easily they can perpetuate stereotypes and contribute to the generation of biased content. With a focus on gender and professional bias, this work examines in which manner LLMs shape responses to ungendered prompts, contributing to biased outputs. This analysis uses a structured experimental method, giving different prompts involving three different professional job combinations, which are also characterized by a hierarchical relationship. This study uses Italian, a language with extensive grammatical gender differences, to highlight potential limitations in current LLMs' ability to generate objective text in non-English languages. Two popular LLM-based chatbots are examined, namely OpenAI ChatGPT (gpt-4o-mini) and Google Gemini (gemini-1.5-flash). Through APIs, we collected a range of 3600 responses. The results highlight how content generated by LLMs can perpetuate stereotypes. For example, Gemini associated 100% (ChatGPT 97%) of 'she' pronouns to the 'assistant' rather than the 'manager'. The presence of bias in AI-generated text can have significant implications in many fields, such as in the workplaces or in job selections, raising ethical concerns about its use. Understanding these risks is pivotal to developing mitigation strategies and assuring that AI-based systems do not increase social inequalities, but rather contribute to more equitable outcomes. Future research directions include expanding the study to additional chatbots or languages, refining prompt engineering methods or further exploiting a larger experimental base.

Understanding another person's creative output requires a shared language of association. However, when training vision-language models such as CLIP, we rely on web-scraped datasets containing short, predominantly literal, alt-text. In this work, we introduce a method for mining contextualized associations for salient visual elements in an image that can scale to any unlabeled dataset. Given an image, we can use these mined associations to generate high quality creative captions at increasing degrees of abstraction. With our method, we produce a new dataset of visual associations and 1.7m creative captions for the images in MSCOCO. Human evaluation confirms that these captions remain visually grounded while exhibiting recognizably increasing abstraction. Moreover, fine-tuning a visual encoder on this dataset yields meaningful improvements in zero-shot image-text retrieval in two creative domains: poetry and metaphor visualization. We release our dataset, our generation code and our models for use by the broader community.

MikeroweWorks Foundation founder Mike Rowe joins'The Brian Kilmeade Show' to discuss how AI and robots threaten white-collar jobs, as the nation faces a need for blue-collar workers. 'UNDENIABLE': Mike Rowe is sounding the alarm about the future of white and blue-collar jobs, and is urging young Americans to rethink their career choices due to threats from artificial intelligence. 'ALL IN': President Donald Trump is going all in on artificial intelligence, with a top Meta executive voicing strong support for his bold strategy. Speaking at a tech summit in Washington, Trump outlined his vision for a future driven by American innovation and secured by global artificial intelligence leadership. INNOVATION BOOST: Nvidia CEO Jensen Huang said in an interview Wednesday that the Trump administration's artificial intelligence plan is poised to boost innovation and AI deployment in the U.S. IMMINENT CRISIS: OpenAI CEO Sam Altman warned Wall Street executives that bad actors could exploit digital voice ID authentication to defraud consumers by enabling large money transfers, creating what he describes as an imminent fraud crisis. STARGATE OPENS: Oracle and OpenAI have inked an agreement to further develop the Stargate project as part of a broader pledge to expand Artificial Intelligence (AI) infrastructure in the United States.

Computers have taken the crown in chess, Go and poker, but when it comes to competitive coding, humans still have the edge – just. Earlier this month Przemysław Dębiak, a Polish coder and mind sports champion, narrowly clinched a victory over OpenAI's entrant in the AtCoder World Tour Finals 2025, in Tokyo. However, the elite coder, who goes by the online name Psyho, predicts he may be the last human to win the prestigious title due to the incredible pace of technological progress. "That's probable," said Psyho, 41, who previously worked at OpenAI before retiring five years ago. "I would prefer not, mostly because I like these competitions and knowing there's this magical entity that can do it better than me would be a little bit frustrating."

On November 2, 2022, I attended a Google AI event in New York City. One of the themes was responsible AI. As I listened to executives talk about how they aligned their technology with human values, I realized that the malleability of AI models was a double-edged sword. Models could be tweaked to, say, minimize biases, but also to enforce a specific point of view. Governments could demand manipulation to censor unwelcome facts and promote propaganda.

We propose a novel framework that harnesses the power of generative artificial intelligence and copula-based modeling to address two critical challenges in multivariate time-series analysis: delivering accurate predictions and enabling robust anomaly detection. Our method, Copula-based Conformal Anomaly Identification for Multivariate Time-Series (CoCAI), leverages a diffusion-based model to capture complex dependencies within the data, enabling high quality forecasting. The model's outputs are further calibrated using a conformal prediction technique, yielding predictive regions which are statistically valid, i.e., cover the true target values with a desired confidence level. Starting from these calibrated forecasts, robust outlier detection is performed by combining dimensionality reduction techniques with copula-based modeling, providing a statistically grounded anomaly score. CoCAI benefits from an offline calibration phase that allows for minimal overhead during deployment and delivers actionable results rooted in established theoretical foundations. Empirical tests conducted on real operational data derived from water distribution and sewerage systems confirm CoCAI's effectiveness in accurately forecasting target sequences of data and in identifying anomalous segments within them.

Text-to-image (T2I) models have become prevalent across numerous applications, making their robust evaluation against adversarial attacks a critical priority. Continuous access to new and challenging adversarial prompts across diverse domains is essential for stress-testing these models for resilience against novel attacks from multiple vectors. Current techniques for generating such prompts are either entirely authored by humans or synthetically generated. On the one hand, datasets of human-crafted adversarial prompts are often too small in size and imbalanced in their cultural and contextual representation. On the other hand, datasets of synthetically-generated prompts achieve scale, but typically lack the realistic nuances and creative adversarial strategies found in human-crafted prompts. To combine the strengths of both human and machine approaches, we propose Seed2Harvest, a hybrid red-teaming method for guided expansion of culturally diverse, human-crafted adversarial prompt seeds. The resulting prompts preserve the characteristics and attack patterns of human prompts while maintaining comparable average attack success rates (0.31 NudeNet, 0.36 SD NSFW, 0.12 Q16). Our expanded dataset achieves substantially higher diversity with 535 unique geographic locations and a Shannon entropy of 7.48, compared to 58 locations and 5.28 entropy in the original dataset. Our work demonstrates the importance of human-machine collaboration in leveraging human creativity and machine computational capacity to achieve comprehensive, scalable red-teaming for continuous T2I model safety evaluation.

Large Language Model (LLM) agents are increasingly utilized in AI-aided education to support tutoring and learning. Effective communication strategies among LLM agents improve collaborative problem-solving efficiency and facilitate cost-effective adoption in education. However, little research has systematically evaluated the impact of different communication strategies on agents' problem-solving. Our study examines four communication modes, \textit{teacher-student interaction}, \textit{peer-to-peer collaboration}, \textit{reciprocal peer teaching}, and \textit{critical debate}, in a dual-agent, chat-based mathematical problem-solving environment using the OpenAI GPT-4o model. Evaluated on the MATH dataset, our results show that dual-agent setups outperform single agents, with \textit{peer-to-peer collaboration} achieving the highest accuracy. Dialogue acts like statements, acknowledgment, and hints play a key role in collaborative problem-solving. While multi-agent frameworks enhance computational tasks, effective communication strategies are essential for tackling complex problems in AI education.

Generating clinical synthetic text represents an effective solution for common clinical NLP issues like sparsity and privacy. This paper aims to conduct a systematic review on generating synthetic medical free-text by formulating quantitative analysis to three research questions concerning (i) the purpose of generation, (ii) the techniques, and (iii) the evaluation methods. We searched PubMed, ScienceDirect, Web of Science, Scopus, IEEE, Google Scholar, and arXiv databases for publications associated with generating synthetic medical unstructured free-text. We have identified 94 relevant articles out of 1,398 collected ones. A great deal of attention has been given to the generation of synthetic medical text from 2018 onwards, where the main purpose of such a generation is towards text augmentation, assistive writing, corpus building, privacy-preserving, annotation, and usefulness. Transformer architectures were the main predominant technique used to generate the text, especially the GPTs. On the other hand, there were four main aspects of evaluation, including similarity, privacy, structure, and utility, where utility was the most frequent method used to assess the generated synthetic medical text. Although the generated synthetic medical text demonstrated a moderate possibility to act as real medical documents in different downstream NLP tasks, it has proven to be a great asset as augmented, complementary to the real documents, towards improving the accuracy and overcoming sparsity/undersampling issues. Yet, privacy is still a major issue behind generating synthetic medical text, where more human assessments are needed to check for the existence of any sensitive information. Despite that, advances in generating synthetic medical text will considerably accelerate the adoption of workflows and pipeline development, discarding the time-consuming legalities of data transfer.