Three people associated with artificial intelligence server maker Super Micro Computer, including its cofounder, have been charged with helping smuggle at least $2.5bn-worth of United States AI technology to China in violation of export laws, according to the US Department of Justice. US prosecutors did not name Super Micro in the complaint, referring only to a "US manufacturer", but San Jose, California-based Super Micro said it was informed by federal prosecutors of the indictment on Thursday. The Justice Department said it had charged Yih-Shyan Liaw, Ruei-Tsang Chang, and Ting-Wei Sun in an indictment unsealed in federal court in Manhattan on Thursday, on allegations of a complex scheme to send US-made servers through Taiwan to other countries in Southeast Asia, where they were swapped into unmarked boxes and sent on to China. The US has had export restrictions on China for advanced AI chips since 2022. In a release, FBI Assistant Director in Charge James Barnacle said the defendants used fabricated documents, staged bogus equipment to pass audit inventories, and used a pass-through company to conceal their misconduct and true clientele list.

The data leak triggered a major internal security alert inside Meta. The data leak triggered a major internal security alert inside Meta. Fri 20 Mar 2026 02.00 EDTLast modified on Fri 20 Mar 2026 03.03 EDT An AI agent instructed an engineer to take actions that exposed a large amount of Meta's sensitive data to some of its employees, in the latest example of AI causing upheaval in a large tech company. The leak, which Meta confirmed, happened when an employee asked for guidance on an engineering problem on an internal forum. An AI agent responded with a solution, which the employee implemented - causing a large amount of sensitive user and company data to be exposed to its engineers for two hours.

Two robots box while German Chancellor Friedrich Merz visits Unitree Robotics in Zhejiang Province, China. In the exhibition hall at Unitree Robotics in Hangzhou, Friedrich Merz smiled and applauded the martial arts display by a platoon of humanoid warriors. But when a robot boxer advanced toward him, punching the air with its red-gloved fists, the German chancellor flinched, a look of alarm crossing his face as he appeared to realize the danger posed by an autonomous fighting machine. It was also a moment that crystallized for Merz the power of China's technology, according to a person familiar with his thinking. He saw it, too, as a sign of how far behind Germany has fallen and how European Union regulation holds back their efforts to catch up, the person said, asking not to be named discussing the chancellor's private views. The trip, last month, has triggered a broader reckoning that is starting to settle in across Europe: Maybe de-risking from China is just too big a task.

A trio linked with a US technology supplier have been charged over a ploy to smuggle American artificial intelligence (AI) chips to China, the Department of Justice said on Thursday. The individuals allegedly conspired to sell billions of dollars' worth of technology to buyers in China by faking documents and using dummy equipment to slip past audits, according to the DOJ. The goods in question included Nvidia-made semiconductors, highly coveted AI chips which are subject to export controls. In August 2025, two Chinese nationals were also arrested and charged with illegally shipping millions of dollars' worth of Nvidia chips to China. The DOJ said in a statement on Thursday that it had arrested US-citizen Yih-Shyan Wally Liaw and Taiwanese citizen Ting-Wei Willy Sun, while Ruei-Tsang Steven Chang, a Taiwanese citizen, remains a fugitive.

The return of BTS is a big deal. In case you were in any doubt, just look at the frenzy surrounding the South Koreans' comeback. On Saturday, the band will kick off a sold-out, 82-date world tour with a free concert in Seoul, which is expected to be attended by more than 250,000 in-person fans and will be live-streamed on Netflix to more than 190 countries. When the tour wraps up in 2027, BTS are expected to have generated more than $1billion in revenue. Some more outlandish estimates suggest they will eclipse the $2billion haul of Taylor Swift's Eras tour.

I witnessed this scene in the Masa Mara park where two young lions were playing with their mother, they were rolling around with her until one of them found himself in a rather uncomfortable and incongruous situation. Breakthroughs, discoveries, and DIY tips sent six days a week. Accidentally running face-first into your mom's butt is funny, no matter your species. The Nikon Comedy Wildlife Awards opened for entries this week and to hype up the 2026 competition, the top wildlife photography competition for not-so-serious animals has released outtakes from last year's contest. We get a caiman with butterfly accessories, a friendly damselfly, and two baboons caught in a compromising position.

Maximum entropy reinforcement learning motivates agents to explore states and actions to maximize the entropy of some distribution, typically by providing additional intrinsic rewards proportional to that entropy function. In this paper, we study intrinsic rewards proportional to the entropy of the discounted distribution of state-action features visited during future time steps. This approach is motivated by two results. First, we show that the expected sum of these intrinsic rewards is a lower bound on the entropy of the discounted distribution of state-action features visited in trajectories starting from the initial states, which we relate to an alternative maximum entropy objective. Second, we show that the distribution used in the intrinsic reward definition is the fixed point of a contraction operator and can therefore be estimated off-policy. Experiments highlight that the new objective leads to improved visitation of features within individual trajectories, in exchange for slightly reduced visitation of features in expectation over different trajectories, as suggested by the lower bound. It also leads to improved convergence speed for learning exploration-only agents. Control performance remains similar across most methods on the considered benchmarks.

Deep learning models have become the dominant approach for multivariate time series anomaly detection (MTSAD), often reporting substantial performance improvements over classical statistical methods. However, these gains are frequently evaluated under heterogeneous thresholding strategies and evaluation protocols, making fair comparisons difficult. This work revisits OmniAnomaly, a widely used stochastic recurrent model for MTSAD, and systematically compares it with a simple linear baseline based on Principal Component Analysis (PCA) on the Server Machine Dataset (SMD). Both methods are evaluated under identical thresholding and evaluation procedures, with experiments repeated across 100 runs for each of the 28 machines in the dataset. Performance is evaluated using Precision, Recall and F1-score at point-level, with and without point-adjustment, and under different aggregation strategies across machines and runs, with the corresponding standard deviations also reported. The results show large variability across machines and show that PCA can achieve performance comparable to OmniAnomaly, and even outperform it when point-adjustment is not applied. These findings question the added value of more complex architectures under current benchmarking practices and highlight the critical role of evaluation methodology in MTSAD research.

Standard decoding strategies for text generation, including top-k, nucleus sampling, and contrastive search, select tokens based on likelihood, restricting selection to high-probability regions. Human language production operates differently: tokens are chosen for communicative appropriateness rather than statistical frequency. This mismatch creates a truncation blind spot: contextually appropriate but statistically rare tokens remain accessible to humans yet unreachable by likelihood-based decoding. We hypothesize this contributes to the detectability of machine-generated text. Analyzing over 1.8 million texts across eight language models, five decoding strategies, and 53 hyperparameter configurations, we find that 8-18% of human-selected tokens fall outside typical truncation boundaries. Simple classifiers trained on predictability and lexical diversity achieve remarkable detection rates. Crucially, neither model scale nor architecture correlates strongly with detectability; truncation parameters account for most variance. Configurations achieving low detectability often produce incoherent text, indicating that evading detection and producing natural text are distinct objectives. These findings suggest detectability is enhanced by likelihood-based token selection, not merely a matter of model capability.

When working with real-world insurance data, practitioners often encounter challenges during the data preparation stage that can undermine the statistical validity and reliability of downstream modeling. This study illustrates that conventional data preparation procedures such as random train-test partitioning, often yield unreliable and unstable results when confronted with highly imbalanced insurance loss data. To mitigate these limitations, we propose a novel data preparation framework leveraging two recent statistical advancements: support points for representative data splitting to ensure distributional consistency across partitions, and the Chatterjee correlation coefficient for initial, non-parametric feature screening to capture feature relevance and dependence structure. We further integrate these theoretical advances into a unified, efficient framework that also incorporates missing-data handling, and embed this framework within our custom InsurAutoML pipeline. The performance of the proposed approach is evaluated using both simulated datasets and datasets often cited in the academic literature. Our findings definitively demonstrate that incorporating statistically rigorous data preparation methods not only significantly enhances model robustness and interpretability but also substantially reduces computational resource requirements across diverse insurance loss modeling tasks. This work provides a crucial methodological upgrade for achieving reliable results in high stakes insurance applications.