Recent advances in machine learning and AI, including Generative AI and LLMs, are disrupting technological innovation, product development, and society as a whole. AI's contribution to technology can come from multiple approaches that require access to large training data sets and clear performance evaluation criteria, ranging from pattern recognition and classification to generative models. Yet, AI has contributed less to fundamental science in part because large data sets of high-quality data for scientific practice and model discovery are more difficult to access. Generative AI, in general, and Large Language Models in particular, may represent an opportunity to augment and accelerate the scientific discovery of fundamental deep science with quantitative models. Here we explore and investigate aspects of an AI-driven, automated, closed-loop approach to scientific discovery, including self-driven hypothesis generation and open-ended autonomous exploration of the hypothesis space. Integrating AI-driven automation into the practice of science would mitigate current problems, including the replication of findings, systematic production of data, and ultimately democratisation of the scientific process. Realising these possibilities requires a vision for augmented AI coupled with a diversity of AI approaches able to deal with fundamental aspects of causality analysis and model discovery while enabling unbiased search across the space of putative explanations. These advances hold the promise to unleash AI's potential for searching and discovering the fundamental structure of our world beyond what human scientists have been able to achieve. Such a vision would push the boundaries of new fundamental science rather than automatize current workflows and instead open doors for technological innovation to tackle some of the greatest challenges facing humanity today.

The International AI Safety Report 2026 synthesises the current scientific evidence on the capabilities, emerging risks, and safety of general-purpose AI systems. The report series was mandated by the nations attending the AI Safety Summit in Bletchley, UK. 29 nations, the UN, the OECD, and the EU each nominated a representative to the report's Expert Advisory Panel. Over 100 AI experts contributed, representing diverse perspectives and disciplines. Led by the Report's Chair, these independent experts collectively had full discretion over the report's content.

The first International AI Safety Report comprehensively synthesizes the current evidence on the capabilities, risks, and safety of advanced AI systems. The report was mandated by the nations attending the AI Safety Summit in Bletchley, UK. Thirty nations, the UN, the OECD, and the EU each nominated a representative to the report's Expert Advisory Panel. A total of 100 AI experts contributed, representing diverse perspectives and disciplines. Led by the report's Chair, these independent experts collectively had full discretion over the report's content.

The field of AI is moving too quickly for a single yearly publication to keep pace. Significant changes can occur on a timescale of months, sometimes weeks. This is why we are releasing Key Updates: shorter, focused reports that highlight the most important developments between full editions of the International AI Safety Report. With these updates, we aim to provide policymakers, researchers, and the public with up-to-date information to support wise decisions about AI governance. This first Key Update focuses on areas where especially significant changes have occurred since January 2025: advances in general-purpose AI systems' capabilities, and the implications for several critical risks. New training techniques have enabled AI systems to reason step-by-step and operate autonomously for longer periods, allowing them to tackle more kinds of work. However, these same advances create new challenges across biological risks, cyber security, and oversight of AI systems themselves. The International AI Safety Report is intended to help readers assess, anticipate, and manage risks from general-purpose AI systems. These Key Updates ensure that critical developments receive timely attention as the field rapidly evolves.

This is the Second Key Update to the 2025 International AI Safety Report. The First Key Update (1) discussed developments in the capabilities of general-purpose AI models and systems and associated risks. This Key Update covers how various actors, including researchers, companies, and governments, are approaching risk management and technical mitigations for AI. The past year has seen important developments in AI risk management, including better techniques for training safer models and monitoring their outputs. While this represents tangible progress, significant gaps remain. It is often uncertain how effective current measures are at preventing harms, and effectiveness varies across time and applications. There are many opportunities to further strengthen existing safeguard techniques and to develop new ones. This Key Update provides a concise overview of critical developments in risk management practices and technical risk mitigation since the publication of the 2025 AI Safety Report in January. It highlights where progress is being made and where gaps remain. Above all, it aims to support policymakers, researchers, and the public in navigating a rapidly changing environment, helping them to make informed and timely decisions about the governance of general-purpose AI. Professor Yoshua BengioUniversité de Montréal / LawZero /Mila – Quebec AI Institute & Chair

Since the publication of the first International AI Safety Report, AI capabilities have continued to improve across key domains. New training techniques that teach AI systems to reason step-by-step and inference-time enhancements have primarily driven these advances, rather than simply training larger models. As a result, general-purpose AI systems can solve more complex problems in a range of domains, from scientific research to software development. Their performance on benchmarks that measure performance in coding, mathematics, and answering expert-level science questions has continued to improve, though reliability challenges persist, with systems excelling on some tasks while failing completely on others. These capability improvements also have implications for multiple risks, including risks from biological weapons and cyber attacks. Finally, they pose new challenges for monitoring and controllability. This update examines how AI capabilities have improved since the first Report, then focuses on key risk areas where substantial new evidence warrants updated assessments.

This second update to the 2025 International AI Safety Report assesses new developments in general-purpose AI risk management over the past year. It examines how researchers, public institutions, and AI developers are approaching risk management for general-purpose AI. In recent months, for example, three leading AI developers applied enhanced safeguards to their new models, as their internal pre-deployment testing could not rule out the possibility that these models could be misused to help create biological weapons. Beyond specific precautionary measures, there have been a range of other advances in techniques for making AI models and systems more reliable and resistant to misuse. These include new approaches in adversarial training, data curation, and monitoring systems. In parallel, institutional frameworks that operationalise and formalise these technical capabilities are starting to emerge: the number of companies publishing Frontier AI Safety Frameworks more than doubled in 2025, and governments and international organisations have established a small number of governance frameworks for general-purpose AI, focusing largely on transparency and risk assessment.

Typical mobile edge computing infrastructures have to contend with unreliable computing devices at their end-points. The limited resource capacities of mobile edge devices gives rise to frequent contentions, node overloads or failures. This is exacerbated by the strict deadlines of modern applications. To avoid failures, fault-tolerant approaches utilize preemptive migration to transfer active tasks across nodes and prevent nodes running at capacity. However, prior work struggles to dynamically adapt in settings with highly volatile workloads or even accurately detect and diagnose anomalies for optimal remediation. To meet the strict service level objectives of contemporary workloads, there is a need for dynamic fault-tolerant methods that can quickly adapt to changes in edge environments while having parsimonious remediation in the form of preemptive migration to avoid stressing the system network. This work proposes PreGAN, featuring a Generative Adversarial Network (GAN) based approach to predict contentions, pinpoint specific resource types with high chance of overload, and generate migration decisions to proactively avoid system downtime. PreGAN leverages coupled-simulations to train the GAN model at run-time and a few-shot fault classifier to update decisions of an underpinning scheduler. We also extend it to PreGAN+ that also periodically tunes the decision model using semi-supervised training and a Transformer based neural network for low tuning time, albeit with higher memory overheads. Experiments on a Raspberry-Pi based edge environment demonstrate that both models outperform state-of-the-art baselines in fault detection and diagnosis scores by up to 12.5% and 31.2% respectively. This also translates in improvements in Quality of Service against baseline approaches.

This is the interim publication of the first International Scientific Report on the Safety of Advanced AI. The report synthesises the scientific understanding of general-purpose AI – AI that can perform a wide variety of tasks – with a focus on understanding and managing its risks. A diverse group of 75 AI experts contributed to this report, including an international Expert Advisory Panel nominated by 30 countries, the EU, and the UN. Led by the Chair, these independent experts collectively had full discretion over the report's content.

Social dilemmas are situations of inter-dependent decision making in which individual rationality can lead to outcomes with poor social qualities. The ubiquity of social dilemmas in social, biological, and computational systems has generated substantial research across these diverse disciplines into the study of mechanisms for avoiding deficient outcomes by promoting and maintaining mutual cooperation. Much of this research is focused on studying how individuals faced with a dilemma can learn to cooperate by adapting their behaviours according to their past experience. In particular, three types of learning approaches have been studied: evolutionary game-theoretic learning, reinforcement learning, and best-response learning. This article is a comprehensive integrated survey of these learning approaches in the context of dilemma games. We formally introduce dilemma games and their inherent challenges. We then outline the three learning approaches and, for each approach, provide a survey of the solutions proposed for dilemma resolution. Finally, we provide a comparative summary and discuss directions in which further research is needed.

The emergence of latency-critical AI applications has been supported by the evolution of the edge computing paradigm. However, edge solutions are typically resource-constrained, posing reliability challenges due to heightened contention for compute capacities and faulty application behavior in the presence of overload conditions. Although a large amount of generated log data can be mined for fault prediction, labeling this data for training is a manual process and thus a limiting factor for automation. Due to this, many companies resort to unsupervised fault-tolerance models. Yet, failure models of this kind can incur a loss of accuracy when they need to adapt to non-stationary workloads and diverse host characteristics. Thus, we propose a novel modeling approach, DeepFT, to proactively avoid system overloads and their adverse effects by optimizing the task scheduling decisions. DeepFT uses a deep-surrogate model to accurately predict and diagnose faults in the system and co-simulation based self-supervised learning to dynamically adapt the model in volatile settings. Experimentation on an edge cluster shows that DeepFT can outperform state-of-the-art methods in fault-detection and QoS metrics. Specifically, DeepFT gives the highest F1 scores for fault-detection, reducing service deadline violations by up to 37% while also improving response time by up to 9%.

Modern large-scale computing systems distribute jobs into multiple smaller tasks which execute in parallel to accelerate job completion rates and reduce energy consumption. However, a common performance problem in such systems is dealing with straggler tasks that are slow running instances that increase the overall response time. Such tasks can significantly impact the system's Quality of Service (QoS) and the Service Level Agreements (SLA). To combat this issue, there is a need for automatic straggler detection and mitigation mechanisms that execute jobs without violating the SLA. Prior work typically builds reactive models that focus first on detection and then mitigation of straggler tasks, which leads to delays. Other works use prediction based proactive mechanisms, but ignore heterogeneous host or volatile task characteristics. In this paper, we propose a Straggler Prediction and Mitigation Technique (START) that is able to predict which tasks might be stragglers and dynamically adapt scheduling to achieve lower response times. Our technique analyzes all tasks and hosts based on compute and network resource consumption using an Encoder Long-Short-Term-Memory (LSTM) network. The output of this network is then used to predict and mitigate expected straggler tasks. This reduces the SLA violation rate and execution time without compromising QoS. Specifically, we use the CloudSim toolkit to simulate START in a cloud environment and compare it with state-of-the-art techniques (IGRU-SD, SGC, Dolly, GRASS, NearestFit and Wrangler) in terms of QoS parameters such as energy consumption, execution time, resource contention, CPU utilization and SLA violation rate. Experiments show that START reduces execution time, resource contention, energy and SLA violations by 13%, 11%, 16% and 19%, respectively, compared to the state-of-the-art approaches.

We consider an outsourcing problem where a software agent procures multiple ser-vices from providers with uncertain reliabilities to complete a computational task before a strict deadline. The service consumer's goal is to design an outsourcing strategy (defining which services to procure and when) so as to maximize a specific objective function. This objective function can be different based on the consumer's nature; a socially-focused con-sumer often aims to maximize social welfare, while a self-interested consumer often aims to maximize its own utility. However, in both cases, the objective function depends on the providers' execution costs, which are privately held by the self-interested providers and hence may be misreported to influence the consumer's decisions. For such settings, we develop a unified approach to design truthful procurement auctions that can be used by both socially-focused and, separately, self-interested consumers. This approach benefits from our proposed weighted threshold payment scheme which pays the provably minimum amount to make an auction with a monotone outsourcing strategy incentive compatible. This payment scheme can handle contingent outsourcing plans, where additional procure-ment happens gradually over time and only if the success probability of the already hired providers drops below a time-dependent threshold. Using a weighted threshold payment scheme, we design two procurement auctions that maximize, as well as two low-complexity heuristic-based auctions that approximately maximize, the consumer's expected utility and expected social welfare, respectively. We demonstrate the effectiveness and strength of our proposed auctions through both game-theoretical and empirical analysis.

The operational cost of a cloud computing platform is one of the most significant Quality of Service (QoS) criteria for schedulers, crucial to keep up with the growing computational demands. Several data-driven deep neural network (DNN)-based schedulers have been proposed in recent years that outperform alternative approaches by providing scalable and effective resource management for dynamic workloads. However, state-of-the-art schedulers rely on advanced DNNs with high computational requirements, implying high scheduling costs. In non-stationary contexts, the most sophisticated schedulers may not always be required, and it may be sufficient to rely on low-cost schedulers to temporarily save operational costs. In this work, we propose MetaNet, a surrogate model that predicts the operational costs and scheduling overheads of a large number of DNNbased schedulers and chooses one on-the-fly to jointly optimize job scheduling and execution costs. This facilitates improvements in execution costs, energy usage and service level agreement violations of up to 11%, 43% and 13% compared to the state-of-the-art methods.

The rise of distributed cloud computing technologies has been pivotal for the large-scale adoption of Artificial Intelligence (AI) based applications for high fidelity and scalable service delivery. Systematic resource management is central in maintaining optimal Quality of Service (QoS) in cloud platforms and is divided into three fundamental types: resource provisioning, AI model deployment and workload placement. To exploit the synergy among these decision types, it becomes imperative to concurrently design (co-design) the provisioning, deployment and placement decisions for optimal QoS. As users and cloud service providers shift to non-stationary AI-based workloads, frequent decision making imposes severe time constraints on the resource management models. Existing AI-based solutions often optimize decision types independently and tend to ignore the dependencies across various system performance aspects such as energy consumption and CPU utilization, making them perform poorly in large-scale cloud systems. To address this, we propose a novel method, called SciNet, that leverages a co-simulated digital-twin of the infrastructure to capture inter-metric dependencies and accurately estimate QoS scores. To avoid expensive simulation overheads at test time, SciNet trains a neural network based imitation learner that aims to mimic an oracle, which takes optimal decisions based on co-simulated QoS estimates. Offline model training and online decision making based on the imitation learner, enables SciNet to take optimal decisions while being time-efficient. Experiments with real-life AI-based benchmark applications on a public cloud testbed show that SciNet gives up to 48% lower execution cost, 79% higher inference accuracy, 71% lower energy consumption and 56% lower response times compared to the current state-of-the-art methods.

In recent years, the landscape of computing paradigms has witnessed a gradual yet remarkable shift from monolithic computing to distributed and decentralized paradigms such as Internet of Things (IoT), Edge, Fog, Cloud, and Serverless. The frontiers of these computing technologies have been boosted by shift from manually encoded algorithms to Artificial Intelligence (AI)-driven autonomous systems for optimum and reliable management of distributed computing resources. Prior work focuses on improving existing systems using AI across a wide range of domains, such as efficient resource provisioning, application deployment, task placement, and service management. This survey reviews the evolution of data-driven AI-augmented technologies and their impact on computing systems. We demystify new techniques and draw key insights in Edge, Fog and Cloud resource management-related uses of AI methods and also look at how AI can innovate traditional applications for enhanced Quality of Service (QoS) in the presence of a continuum of resources. We present the latest trends and impact areas such as optimizing AI models that are deployed on or for computing systems. We layout a roadmap for future research directions in areas such as resource management for QoS optimization and service reliability. Finally, we discuss blue-sky ideas and envision this work as an anchor point for future research on AI-driven computing systems.

Intelligent Virtual Machine (VM) provisioning is central to cost and resource efficient computation in cloud computing environments. As bootstrapping VMs is time-consuming, a key challenge for latency-critical tasks is to predict future workload demands to provision VMs proactively. However, existing AI-based solutions tend to not holistically consider all crucial aspects such as provisioning overheads, heterogeneous VM costs and Quality of Service (QoS) of the cloud system. To address this, we propose a novel method, called CILP, that formulates the VM provisioning problem as two sub-problems of prediction and optimization, where the provisioning plan is optimized based on predicted workload demands. CILP leverages a neural network as a surrogate model to predict future workload demands with a co-simulated digital-twin of the infrastructure to compute QoS scores. We extend the neural network to also act as an imitation learner that dynamically decides the optimal VM provisioning plan. A transformer based neural model reduces training and inference overheads while our novel two-phase decision making loop facilitates in making informed provisioning decisions. Crucially, we address limitations of prior work by including resource utilization, deployment costs and provisioning overheads to inform the provisioning decisions in our imitation learning framework. Experiments with three public benchmarks demonstrate that CILP gives up to 22% higher resource utilization, 14% higher QoS scores and 44% lower execution costs compared to the current online and offline optimization based state-of-the-art methods.

Intrusion research frequently collects data on attack techniques currently employed and their potential symptoms. This includes deploying honeypots, logging events from existing devices, employing a red team for a sample attack campaign, or simulating system activity. However, these observational studies do not clearly discern the cause-and-effect relationships between the design of the environment and the data recorded. Neglecting such relationships increases the chance of drawing biased conclusions due to unconsidered factors, such as spurious correlations between features and errors in measurement or classification. In this paper, we present the theory and empirical data on methods that aim to discover such causal relationships efficiently. Our adaptive design (AD) is inspired by the clinical trial community: a variant of a randomized control trial (RCT) to measure how a particular ``treatment'' affects a population. To contrast our method with observational studies and RCT, we run the first controlled and adaptive honeypot deployment study, identifying the causal relationship between an ssh vulnerability and the rate of server exploitation. We demonstrate that our AD method decreases the total time needed to run the deployment by at least 33%, while still confidently stating the impact of our change in the environment. Compared to an analogous honeypot study with a control group, our AD requests 17% fewer honeypots while collecting 19% more attack recordings than an analogous honeypot study with a control group.

We consider a procurement problem where a software agent procures multiple services from self-interested providers with private costs and uncertain reliabilities to complete a budget-limited task before a strict deadline. Over the last decade, several truthful budget-feasible procurement mechanisms have been developed to extract the true cost information from strategic providers. Most of these mechanisms have focused on maximizing the procurer's value (e.g., the task's success probability), and hence procuring as many services as the budget allows, even if the returned benefit is lower than the incurred cost. In this paper, however, we focus on the more realistic objective of balancing the cost-benefit tradeoff and propose a novel approach for designing budget-feasible mechanisms that invoke services gradually over time and whenever they are cost-optimal. A major barrier to achieving this goal was the strong dependencies among the decision variables caused by budget constraints. We overcome this barrier by proposing a conservative decomposable approximation to budget constraints. This is the first such approximation technique, which opens a path toward designing budget-feasible mechanisms for contingent planning problems.