Despite Artificial Intelligence (AI) transformative potential for public sector services, decision-making, and administrative efficiency, adoption remains uneven due to complex technical, organizational, and institutional challenges. Responsible AI frameworks emphasize fairness, accountability, and transparency, aligning with principles of trustworthy AI and fair AI, yet remain largely aspirational, overlooking technical and institutional realities, especially foundational data and governance. This study addresses this gap by developing a taxonomy of data-related challenges to responsible AI adoption in government. Based on a systematic review of 43 studies and 21 expert evaluations, the taxonomy identifies 13 key challenges across technological, organizational, and environmental dimensions, including poor data quality, limited AI-ready infrastructure, weak governance, misalignment in human-AI decision-making, economic and environmental sustainability concerns. Annotated with institutional pressures, the taxonomy serves as a diagnostic tool to surface 'symptoms' of high-risk AI deployment and guides policymakers in building the institutional and data governance conditions necessary for responsible AI adoption.

Building agents, systems that perceive and act upon their environment with a degree of autonomy, has long been a focus of AI research. This pursuit has recently become vastly more practical with the emergence of large language models (LLMs) capable of using natural language to integrate information, follow instructions, and perform forms of "reasoning" and planning across a wide range of tasks. With its multimodal data streams and orchestrated workflows spanning multiple systems, radiology is uniquely suited to benefit from agents that can adapt to context and automate repetitive yet complex tasks. In radiology, LLMs and their multimodal variants have already demonstrated promising performance for individual tasks such as information extraction and report summarization. However, using LLMs in isolation underutilizes their potential to support complex, multi-step workflows where decisions depend on evolving context from multiple information sources. Equipping LLMs with external tools and feedback mechanisms enables them to drive systems that exhibit a spectrum of autonomy, ranging from semi-automated workflows to more adaptive agents capable of managing complex processes. This review examines the design of such LLM-driven agentic systems, highlights key applications, discusses evaluation methods for planning and tool use, and outlines challenges such as error cascades, tool-use efficiency, and health IT integration.

Recent advancements in language agents have led to significant improvements in multi-hop reasoning tasks. However, existing approaches often struggle with handling open-domain problems, which require massive information retrieval due to their reliance on a fixed sequence of actions. To address this, we propose Feedback-Guided Dynamic Interactive Planning (FGDIP), a novel framework tailored to enhance reasoning in LLMs by utilizing dynamic and adaptive strategies for information exploration in open-domain multi-hop reasoning tasks. Our approach begins by identifying key entities relevant to the problem, which serve as the initial nodes in the reasoning process. From these initial nodes, we then generate reasoning child nodes with the process being refined through a combination of historical error analysis and real-time feedback, which allows the framework to dynamically adjust and optimize its reasoning strategies. By integrating depth-first search with an innovative node generation technique, our framework adapts based on both prior error paths and concurrently generated nodes at the same hierarchical level. This dynamic strategy effectively expands the search space while ensuring the reasoning process systematically converges toward accurate solutions. Experimental results show that FGDIP achieved up to 54.47% F1 score on the HotpotQA dataset and 70.05% on the StrategyQA dataset, surpassing the best baseline by 5.03% and 7.25% respectively, highlighting its versatility and potential to enhance language agents in multi-hop reasoning tasks.

As governments move to regulate AI, there is growing interest in using Large Language Models (LLMs) to assess whether or not an AI system complies with a given AI Regulation (AIR). However, there is presently no way to benchmark the performance of LLMs at this task. To fill this void, we introduce AIReg-Bench: the first benchmark dataset designed to test how well LLMs can assess compliance with the EU AI Act (AIA). We created this dataset through a two-step process: (1) by prompting an LLM with carefully structured instructions, we generated 120 technical documentation excerpts (samples), each depicting a fictional, albeit plausible, AI system - of the kind an AI provider might produce to demonstrate their compliance with AIR; (2) legal experts then reviewed and annotated each sample to indicate whether, and in what way, the AI system described therein violates specific Articles of the AIA. The resulting dataset, together with our evaluation of whether frontier LLMs can reproduce the experts' compliance labels, provides a starting point to understand the opportunities and limitations of LLM-based AIR compliance assessment tools and establishes a benchmark against which subsequent LLMs can be compared. The dataset and evaluation code are available at https://github.com/camlsys/aireg-bench.

Effective analysis of cybersecurity and threat intelligence data demands language models that can interpret specialized terminology, complex document structures, and the interdependence of natural language and source code. Encoder-only transformer architectures provide efficient and robust representations that support critical tasks such as semantic search, technical entity extraction, and semantic analysis, which are key to automated threat detection, incident triage, and vulnerability assessment. However, general-purpose language models often lack the domain-specific adaptation required for high precision. We present SecureBERT 2.0, an enhanced encoder-only language model purpose-built for cybersecurity applications. Leveraging the ModernBERT architecture, SecureBERT 2.0 introduces improved long-context modeling and hierarchical encoding, enabling effective processing of extended and heterogeneous documents, including threat reports and source code artifacts. Pretrained on a domain-specific corpus more than thirteen times larger than its predecessor, comprising over 13 billion text tokens and 53 million code tokens from diverse real-world sources, SecureBERT 2.0 achieves state-of-the-art performance on multiple cybersecurity benchmarks. Experimental results demonstrate substantial improvements in semantic search for threat intelligence, semantic analysis, cybersecurity-specific named entity recognition, and automated vulnerability detection in code within the cybersecurity domain.

The 2025 OpenAI Preparedness Framework does not guarantee any AI risk mitigation practices: a proof-of-concept for affordance analyses of AI safety policies. Abstract Prominent AI companies are producing'safety frameworks' as a type of voluntary self-governance. These statements purport to establish risk thresholds and safety procedures for the development and deployment of highly capable AI. Understanding which AI risks are covered and what actions are allowed, refused, demanded, encouraged, or discouraged by these statements is vital for assessing how these frameworks actually govern AI development and deployment. We draw on affordance theory to analyse the OpenAI'Preparedness Framework Version 2' (April 2025) using the Mechanisms & Conditions model of affordances and the MIT AI Risk Repository. We find that this safety policy requests evaluation of a small minority of AI risks, encourages deployment of systems with'Medium' capabilities for unintentionally enabling'severe harm' (which OpenAI defines as >1000 deaths or >$100B in damages), and allows OpenAI's CEO to deploy even more dangerous capabilities. These findings suggest that effective mitigation of AI risks requires more robust governance interventions beyond current industry self-regulation. Our affordance analysis provides a replicable method for evaluating what safety frameworks actually permit versus what they claim.

Adapting billion-parameter language models to a downstream task is still costly, even with parameter-efficient fine-tuning (PEFT). We re-cast task adaptation as output-distribution alignment: the objective is to steer the output distribution toward the task distribution directly during decoding rather than indirectly through weight updates. Building on this view, we introduce Steering Vector Decoding (SVDecode), a lightweight, PEFT-compatible, and theoretically grounded method. We start with a short warm-start fine-tune and extract a task-aware steering vector from the Kullback-Leibler (KL) divergence gradient between the output distribution of the warm-started and pre-trained models. This steering vector is then used to guide the decoding process to steer the model's output distribution towards the task distribution. We theoretically prove that SVDecode is first-order equivalent to the gradient step of full fine-tuning and derive a globally optimal solution for the strength of the steering vector. Across three tasks and nine benchmarks, SVDecode paired with four standard PEFT methods improves multiple-choice accuracy by up to 5 percentage points and open-ended truthfulness by 2 percentage points, with similar gains (1-2 percentage points) on commonsense datasets without adding trainable parameters beyond the PEFT adapter. SVDecode thus offers a lightweight, theoretically grounded path to stronger task adaptation for large language models.

This paper presents ASTREA, the first agentic system executed on flight-heritage hardware (TRL 9) for autonomous spacecraft operations, with on-orbit operation aboard the International Space Station (ISS). Using thermal control as a representative use case, we integrate a resource-constrained Large Language Model (LLM) agent with a reinforcement learning controller in an asynchronous architecture tailored for space-qualified platforms. Ground experiments show that LLM-guided supervision improves thermal stability and reduces violations, confirming the feasibility of combining semantic reasoning with adaptive control under hardware constraints. On-orbit validation aboard the ISS initially faced challenges due to inference latency misaligned with the rapid thermal cycles of Low Earth Orbit (LEO) satellites. Synchronization with the orbit length successfully surpassed the baseline with reduced violations, extended episode durations, and improved CPU utilization. These findings demonstrate the potential for scalable agentic supervision architectures in future autonomous spacecraft.

Large language models with Mixture-of-Experts (MoE) architectures achieve efficiency and scalability, yet their routing mechanisms introduce safety alignment challenges insufficiently addressed by techniques developed for dense models. In this work, the MoE-specific safety risk of positional vulnerability-that safety-aligned behaviors rely on specific expert modules-is formalized and systematically analyzed. An analytical framework, SAFEx, is presented to robustly identify, characterize, and validate safety-critical experts via a stability-based expert selection procedure, and to decompose them into two functional groups: the Harmful Content Detection Group (HCDG), which specializes in identifying and recognizing harmful content within user inputs, and the Harmful Response Control Group (HRCG), which specializes in controlling and enforcing model behaviors to generate appropriate safety responses. Expert-level interventions are conducted to probe causality and to test mitigation. Targeted masking of SAFEx-selected experts reveals that safety behavior is highly concentrated. On Qwen3-30B-A3B, configured with 48 MoE-FFN layers and 128 experts per layer under top-8 routing (48x128=6,144 experts in total), disabling 12 selected experts reduces the refusal rate by 22%. In addition, lightweight adaptation is performed using LoRA under three configurations-the HRCG, the union of HCDG and HRCG, and all experts-and the resulting updates are composed through negative weight merging targeted at the HRCG, leading to improved refusal under adversarial prompts without full-model retraining. These results establish positional vulnerability as a distinct MoE-specific safety challenge and provide a practical, compute-efficient pathway for expert-level safety interventions within routed architectures.

Federated Learning (FL) offers a promising approach for training clinical AI models without centralizing sensitive patient data. However, its real-world adoption is hindered by challenges related to privacy, resource constraints, and compliance. Existing Differential Privacy (DP) approaches often apply uniform noise, which disproportionately degrades model performance, even among well-compliant institutions. In this work, we propose a novel compliance-aware FL framework that enhances DP by adaptively adjusting noise based on quantifiable client compliance scores. Additionally, we introduce a compliance scoring tool based on key healthcare and security standards to promote secure, inclusive, and equitable participation across diverse clinical settings. Extensive experiments on public datasets demonstrate that integrating under-resourced, less compliant clinics with highly regulated institutions yields accuracy improvements of up to 15% over traditional FL. This work advances FL by balancing privacy, compliance, and performance, making it a viable solution for real-world clinical workflows in global healthcare.