The US administration's targeting of academic research and international students is a "great gift" to China in the race to compete on artificial intelligence, former OpenAI board member Helen Toner has said. The director of strategy at Georgetown's Center for Security and Emerging Technology (CSET) joined the board of OpenAI in 2021 after a career studying AI and the relationship between the United States and China. Toner, a 33-year-old University of Melbourne graduate, was on the board for two years until a falling out with founder Sam Altman in 2023. Altman was fired by the board over claims that he was not "consistently candid" in his communications and the board did not have confidence in Altman's ability to lead. The chaotic months that followed saw Altman fired and then re-hired with three members of the board, including Toner, ousted instead.

This will not be ready before the next king's speech, and is likely to trigger concerns about delays to regulating the technology. The date for the next king's speech has not been set but several sources said it could take place in May 2026. Labour had originally planned to introduce a short, narrowly drafted AI bill within months of entering office that would have been focused on large language models, such as ChatGPT. The legislation would have required companies to hand over their models for testing by the UK's AI Security Institute. It was intended to address concerns that AI models could become so advanced that they posed a risk to humanity.

Sam Altman co-founded Tools for Humanity in 2019 as part of a suite of companies he believed would reshape the world. Once the tech he was developing at OpenAI passed a certain level of intelligence, he reasoned, it would mark the end of one era on the Internet and the beginning of another, in which AI became so advanced, so human-like, that you would no longer be able to tell whether what you read, saw, or heard online came from a real person. When that happened, Altman imagined, we would need a new kind of online infrastructure: a human-verification layer for the Internet, to distinguish real people from the proliferating number of bots and AI "agents." TIME Correspondent Billy Perrigo explains the solution that Altman came up with - a mysterious device called the Orb.

Earlier this week, OpenAI announced that ChatGPT can now be connected to more apps and services, allowing you to pull your data from those sources in real time. Newly connectable sources that were explicitly mentioned include Outlook, Teams, Google Drive, Gmail, and Linear. ChatGPT can now connect to more internal sources & pull in real-time context--keeping existing user-level permissions. These new connections are only available for paid ChatGPT Plus, Pro, Team, Enterprise, and Edu users. Furthermore, the feature is excluded for users in the European Economic Area (EEA), Switzerland, and UK.

With large language models' (LLMs) continued successes in achieving high scores on various benchmarks [Ope25; Dee+25; Ant25; Tea+25], there still remains a question of how well these scores translate into real-world performance. In the real world there are often questions with no solutions because problems are either underdetermined, overdetermined or simply ill-posed. The ability to ask right questions (and filter out the fluff before answering them) is arguably no less valuable than the ability to answer the questions with given answers. This is in a stark contrast with current benchmark evaluations (and training [Dee+25; Lam+25]) approach, which are supposed to be crafted carefully enough to have at least a single unambiguous solution. We propose that the models should be systematically tested for the existence of such a "bias", which, if present, might translate into models always trying to find a solution, even when the right thing is to say that the question is ill-posed, and in turn sabotage the potential for success of (semi-)autonomy envisioned for the agents built upon these models.

Recently, Large Language Models (LLMs) and Vision Large Language Models (VLLMs) have demonstrated impressive performance as agents across various tasks while data scarcity and label noise remain significant challenges in computer vision tasks, such as object detection and instance segmentation. A common solution for resolving these issues is to generate synthetic data. However, current synthetic data generation methods struggle with issues, such as multiple objects per mask, inaccurate segmentation, and incorrect category labels, limiting their effectiveness. To address these issues, we introduce Gen-n-Val, a novel agentic data generation framework that leverages Layer Diffusion (LD), LLMs, and VLLMs to produce high-quality, single-object masks and diverse backgrounds. Gen-n-Val consists of two agents: (1) The LD prompt agent, an LLM, optimizes prompts for LD to generate high-quality foreground instance images and segmentation masks. These optimized prompts ensure the generation of single-object synthetic data with precise instance masks and clean backgrounds. (2) The data validation agent, a VLLM, which filters out low-quality synthetic instance images. The system prompts for both agents are refined through TextGrad. Additionally, we use image harmonization to combine multiple instances within scenes. Compared to state-of-the-art synthetic data approaches like MosaicFusion, our approach reduces invalid synthetic data from 50% to 7% and improves performance by 1% mAP on rare classes in COCO instance segmentation with YOLOv9c and YOLO11m. Furthermore, Gen-n-Val shows significant improvements (7. 1% mAP) over YOLO-Worldv2-M in open-vocabulary object detection benchmarks with YOLO11m. Moreover, Gen-n-Val improves the performance of YOLOv9 and YOLO11 families in instance segmentation and object detection.

The progress of AI systems such as large language models (LLMs) raises increasingly pressing concerns about their safe deployment. This paper examines the value alignment problem for LLMs, arguing that current alignment strategies are fundamentally inadequate to prevent misuse. Despite ongoing efforts to instill norms such as helpfulness, honesty, and harmlessness in LLMs through fine-tuning based on human preferences, they remain vulnerable to adversarial attacks that exploit conflicts between these norms. I argue that this vulnerability reflects a fundamental limitation of existing alignment methods: they reinforce shallow behavioral dispositions rather than endowing LLMs with a genuine capacity for normative deliberation. Drawing from on research in moral psychology, I show how humans' ability to engage in deliberative reasoning enhances their resilience against similar adversarial tactics. LLMs, by contrast, lack a robust capacity to detect and rationally resolve normative conflicts, leaving them susceptible to manipulation; even recent advances in reasoning-focused LLMs have not addressed this vulnerability. This ``shallow alignment'' problem carries significant implications for AI safety and regulation, suggesting that current approaches are insufficient for mitigating potential harms posed by increasingly capable AI systems.

This paper introduces MMRefine, a MultiModal Refinement benchmark designed to evaluate the error refinement capabilities of Multimodal Large Language Models (MLLMs). As the emphasis shifts toward enhancing reasoning during inference, MMRefine provides a framework that evaluates MLLMs' abilities to detect and correct errors across six distinct scenarios beyond just comparing final accuracy before and after refinement. Furthermore, the benchmark analyzes the refinement performance by categorizing errors into six error types. Experiments with various open and closed MLLMs reveal bottlenecks and factors impeding refinement performance, highlighting areas for improvement in effective reasoning enhancement. Our code and dataset are publicly available at https://github.com/naver-ai/MMRefine.

Recent advancements in large language models (LLMs) have spurred interest in robotic navigation that incorporates complex spatial, mathematical, and conditional constraints from natural language into the planning problem. Such constraints can be informal yet highly complex, making it challenging to translate into a formal description that can be passed on to a planning algorithm. In this paper, we propose STPR, a constraint generation framework that uses LLMs to translate constraints (expressed as instructions on ``what not to do'') into executable Python functions. STPR leverages the LLM's strong coding capabilities to shift the problem description from language into structured and transparent code, thus circumventing complex reasoning and avoiding potential hallucinations. We show that these LLM-generated functions accurately describe even complex mathematical constraints, and apply them to point cloud representations with traditional search algorithms. Experiments in a simulated Gazebo environment show that STPR ensures full compliance across several constraints and scenarios, while having short runtimes. We also verify that STPR can be used with smaller, code-specific LLMs, making it applicable to a wide range of compact models at low inference cost.

Scientific Natural Language Inference (NLI) is the task of predicting the semantic relation between a pair of sentences extracted from research articles. Existing datasets for this task are derived from various computer science (CS) domains, whereas non-CS domains are completely ignored. In this paper, we introduce a novel evaluation benchmark for scientific NLI, called MISMATCHED. The new MISMATCHED benchmark covers three non-CS domains-PSYCHOLOGY, ENGINEERING, and PUBLIC HEALTH, and contains 2,700 human annotated sentence pairs. We establish strong baselines on MISMATCHED using both Pre-trained Small Language Models (SLMs) and Large Language Models (LLMs). Our best performing baseline shows a Macro F1 of only 78.17% illustrating the substantial headroom for future improvements. In addition to introducing the MISMATCHED benchmark, we show that incorporating sentence pairs having an implicit scientific NLI relation between them in model training improves their performance on scientific NLI. We make our dataset and code publicly available on GitHub.