The recent development of powerful AI systems has highlighted the need for robust risk management frameworks in the AI industry. Although companies have begun to implement safety frameworks, current approaches often lack the systematic rigor found in other high-risk industries. This paper presents a comprehensive risk management framework for the development of frontier AI that bridges this gap by integrating established risk management principles with emerging AI-specific practices. The framework consists of four key components: (1) risk identification (through literature review, open-ended red-teaming, and risk modeling), (2) risk analysis and evaluation using quantitative metrics and clearly defined thresholds, (3) risk treatment through mitigation measures such as containment, deployment controls, and assurance processes, and (4) risk governance establishing clear organizational structures and accountability. Drawing from best practices in mature industries such as aviation or nuclear power, while accounting for AI's unique challenges, this framework provides AI developers with actionable guidelines for implementing robust risk management. The paper details how each component should be implemented throughout the life-cycle of the AI system - from planning through deployment - and emphasizes the importance and feasibility of conducting risk management work prior to the final training run to minimize the burden associated with it.

We introduce a professionally translated extension of the TruthfulQA benchmark designed to evaluate truthfulness in Basque, Catalan, Galician, and Spanish. Truthfulness evaluations of large language models (LLMs) have primarily been conducted in English. However, the ability of LLMs to maintain truthfulness across languages remains under-explored. Our study evaluates 12 state-of-the-art open LLMs, comparing base and instruction-tuned models using human evaluation, multiple-choice metrics, and LLM-as-a-Judge scoring. Our findings reveal that, while LLMs perform best in English and worst in Basque (the lowest-resourced language), overall truthfulness discrepancies across languages are smaller than anticipated. Furthermore, we show that LLM-as-a-Judge correlates more closely with human judgments than multiple-choice metrics, and that informativeness plays a critical role in truthfulness assessment. Our results also indicate that machine translation provides a viable approach for extending truthfulness benchmarks to additional languages, offering a scalable alternative to professional translation. Finally, we observe that universal knowledge questions are better handled across languages than context- and time-dependent ones, highlighting the need for truthfulness evaluations that account for cultural and temporal variability. Dataset and code are publicly available under open licenses.

We present ExoMiner++, an enhanced deep learning model that builds on the success of ExoMiner to improve transit signal classification in 2-minute TESS data. ExoMiner++ incorporates additional diagnostic inputs, including periodogram, flux trend, difference image, unfolded flux, and spacecraft attitude control data, all of which are crucial for effectively distinguishing transit signals from more challenging sources of false positives. To further enhance performance, we leverage transfer learning from high-quality labeled data from the Kepler space telescope, mitigating the impact of TESS's noisier and more ambiguous labels. ExoMiner++ achieves high accuracy across various classification and ranking metrics, significantly narrowing the search space for follow-up investigations to confirm new planets. To serve the exoplanet community, we introduce new TESS catalogs containing ExoMiner++ classifications and confidence scores for each transit signal. Among the 147,568 unlabeled TCEs, ExoMiner++ identifies 7,330 as planet candidates, with the remainder classified as false positives. These 7,330 planet candidates correspond to 1,868 existing TESS Objects of Interest (TOIs), 69 Community TESS Objects of Interest (CTOIs), and 50 newly introduced CTOIs. 1,797 out of the 2,506 TOIs previously labeled as planet candidates in ExoFOP are classified as planet candidates by ExoMiner++. This reduction in plausible candidates combined with the excellent ranking quality of ExoMiner++ allows the follow-up efforts to be focused on the most likely candidates, increasing the overall planet yield.

In this work-in-progress, we investigate the certification of AI systems, focusing on the practical application and limitations of existing certification catalogues in the light of the AI Act by attempting to certify a publicly available AI system. We aim to evaluate how well current approaches work to effectively certify an AI system, and how publicly accessible AI systems, that might not be actively maintained or initially intended for certification, can be selected and used for a sample certification process. Our methodology involves leveraging the Fraunhofer AI Assessment Catalogue as a comprehensive tool to systematically assess an AI model's compliance with certification standards. We find that while the catalogue effectively structures the evaluation process, it can also be cumbersome and time-consuming to use. We observe the limitations of an AI system that has no active development team anymore and highlighted the importance of complete system documentation. Finally, we identify some limitations of the certification catalogues used and proposed ideas on how to streamline the certification process.

Bathymetry, the study of underwater topography, relies on sonar mapping of submerged structures. These measurements, critical for infrastructure health monitoring, often require expensive instrumentation. The high financial risk associated with sensor damage or vessel loss creates a reluctance to deploy uncrewed surface vessels (USVs) for bathymetry. However, the crewed-boat bathymetry operations, are costly, pose hazards to personnel, and frequently fail to achieve the stable conditions necessary for bathymetry data collection, especially under high currents. Further research is essential to advance autonomous control, navigation, and data processing technologies, with a particular focus on bathymetry. There is a notable lack of accessible hardware platforms that allow for integrated research in both bathymetry-focused autonomous control and navigation, as well as data evaluation and processing. This paper addresses this gap through the design and implementation of two complementary USV systems tailored for uncrewed bathymetry research. This includes a low-cost USV for Navigation And Control research (NAC-USV) and a second, high-end USV equipped with a high-resolution multi-beam sonar and the associated hardware for Bathymetry data quality Evaluation and Post-processing research (BEP-USV). The NAC-USV facilitates the investigation of autonomous, fail-safe navigation and control, emphasizing the stability requirements for high-quality bathymetry data collection while minimizing the risk to equipment. The BEP-USV, which mirrors the NAC-USV hardware, is then used for additional control validation and in-depth exploration of bathymetry data evaluation and post-processing methodologies. We detail the design and implementation of both systems, and open source the design. Furthermore, we demonstrate the system's effectiveness in a range of operational scenarios.

Uncertainty Quantification (UQ) for Natural Language Generation (NLG) is crucial for assessing the performance of Large Language Models (LLMs), as it reveals confidence in predictions, identifies failure modes, and gauges output reliability. Conformal Prediction (CP), a model-agnostic method that generates prediction sets with a specified error rate, has been adopted for UQ in classification tasks, where the size of the prediction set indicates the model's uncertainty. However, when adapting CP to NLG, the sampling-based method for generating candidate outputs cannot guarantee the inclusion of the ground truth, limiting its applicability across a wide range of error rates. To address this, we propose \ourmethod, a method that explicitly adds the ground truth to the candidate outputs and uses logit scores to measure nonconformity. Our experiments with six LLMs on four NLG tasks show that \ourmethod outperforms baseline methods in calibrating error rates and empirical cover rates, offering accurate UQ across a wide range of user-specified error rates.

Machine-generated Text (MGT) detection is crucial for regulating and attributing online texts. While the existing MGT detectors achieve strong performance, they remain vulnerable to simple perturbations and adversarial attacks. To build an effective defense against malicious perturbations, we view MGT detection from a threat modeling perspective, that is, analyzing the model's vulnerability from an adversary's point of view and exploring effective mitigations. To this end, we introduce an adversarial framework for training a robust MGT detector, named GREedy Adversary PromoTed DefendER (GREATER). The GREATER consists of two key components: an adversary GREATER-A and a detector GREATER-D. The GREATER-D learns to defend against the adversarial attack from GREATER-A and generalizes the defense to other attacks. GREATER-A identifies and perturbs the critical tokens in embedding space, along with greedy search and pruning to generate stealthy and disruptive adversarial examples. Besides, we update the GREATER-A and GREATER-D synchronously, encouraging the GREATER-D to generalize its defense to different attacks and varying attack intensities. Our experimental results across 9 text perturbation strategies and 5 adversarial attacks show that our GREATER-D reduces the Attack Success Rate (ASR) by 10.61% compared with SOTA defense methods while our GREATER-A is demonstrated to be more effective and efficient than SOTA attack approaches.

Large language models (LLMs) demonstrate unprecedented capabilities and define the state of the art for almost all natural language processing (NLP) tasks and also for essentially all Language Technology (LT) applications. LLMs can only be trained for languages for which a sufficient amount of pre-training data is available, effectively excluding many languages that are typically characterised as under-resourced. However, there is both circumstantial and empirical evidence that multilingual LLMs, which have been trained using data sets that cover multiple languages (including under-resourced ones), do exhibit strong capabilities for some of these under-resourced languages. Eventually, this approach may have the potential to be a technological off-ramp for those under-resourced languages for which "native" LLMs, and LLM-based technologies, cannot be developed due to a lack of training data. This paper, which concentrates on European languages, examines this idea, analyses the current situation in terms of technology support and summarises related work. The article concludes by focusing on the key open questions that need to be answered for the approach to be put into practice in a systematic way.

Warning: This paper contains potentially offensive and harmful text. Large Reasoning Models (LRMs) have recently extended their powerful reasoning capabilities to safety checks--using chain-of-thought reasoning to decide whether a request should be answered. While this new approach offers a promising route for balancing model utility and safety, its robustness remains underexplored. To address this gap, we introduce Malicious-Educator, a benchmark that disguises extremely dangerous or malicious requests beneath seemingly legitimate educational prompts. Our experiments reveal severe security flaws in popular commercial-grade LRMs, including OpenAI o1/o3, DeepSeek-R1, and Gemini 2.0 Flash Thinking. For instance, although OpenAI's o1 model initially maintains a high refusal rate of about 98%, subsequent model updates significantly compromise its safety; and attackers can easily extract criminal strategies from DeepSeek-R1 and Gemini 2.0 Flash Thinking without any additional tricks. To further highlight these vulnerabilities, we propose Hijacking Chain-of-Thought (H-CoT), a universal and transferable attack method that leverages the model's own displayed intermediate reasoning to jailbreak its safety reasoning mechanism. Under H-CoT, refusal rates sharply decline--dropping from 98% to below 2%--and, in some instances, even transform initially cautious tones into ones that are willing to provide harmful content. We hope these findings underscore the urgent need for more robust safety mechanisms to preserve the benefits of advanced reasoning capabilities without compromising ethical standards.

Meeting summarization suffers from limited high-quality data, mainly due to privacy restrictions and expensive collection processes. We address this gap with FAME, a dataset of 500 meetings in English and 300 in German produced by MIMIC, our new multi-agent meeting synthesis framework that generates meeting transcripts on a given knowledge source by defining psychologically grounded participant profiles, outlining the conversation, and orchestrating a large language model (LLM) debate. A modular post-processing step refines these outputs, mitigating potential repetitiveness and overly formal tones, ensuring coherent, credible dialogues at scale. We also propose a psychologically grounded evaluation framework assessing naturalness, social behavior authenticity, and transcript difficulties. Human assessments show that FAME approximates real-meeting spontaneity (4.5/5 in naturalness), preserves speaker-centric challenges (3/5 in spoken language), and introduces richer information-oriented difficulty (4/5 in difficulty). These findings highlight that FAME is a good and scalable proxy for real-world meeting conditions. It enables new test scenarios for meeting summarization research and other conversation-centric applications in tasks requiring conversation data or simulating social scenarios under behavioral constraints.