In an exclusive interview with WIRED, Paragon Solutions CEO Andrew Boyd reveals the limits of the company's promise to keep bad actors from abusing its powerful espionage tool. Spyware maker Paragon Solutions has long positioned itself as the good guy in an industry seemingly filled with bad ones, vowing to never sell its mobile spyware to authoritarian regimes or ones with poor human rights records. It also promises to cut off any customer caught misusing its products against journalists, dissidents, or other non-legitimate targets. Yet weeks after Paragon, then Israeli-owned, was acquired by the US equity firm AE Industrial Partners in December 2024 and merged with REDLattice--an American offensive cyber firm owned by AE that this week announced plans to go public --WhatsApp alleged that Paragon's Graphite spyware was used to infect the phones of more than 60 individuals in more than 20 countries, including journalists and activists. Most of the targets were not identified, but the University of Toronto's Citizen Lab named two journalists and two activists in Italy. Italian authorities denied misuse . Paragon and its new US owners, despite their zero-tolerance policy for customer abuse of their software, initially declined to comment on the allegations, and reportedly was exploring potential legal action against WhatsApp after the company sent a cease-and-desist letter to Paragon.

Yoshihisa Kainuma, chairman of Minebea Mitsumi, says the company plans to focus on ballooning orders for its ball bearings, used inside AI server fans and humanoid joints. Minebea Mitsumi, one of Japan's biggest corporate dealmakers, is hitting pause on acquisitions to focus on its output of ball bearings, motors and actuators for AI hardware to chase lucrative returns from the emerging technology. The key supplier to Nintendo is eyeing climbing market forecasts for data centers and humanoid robots, and sees "unlimited" upside ahead, according to Chairman Yoshihisa Kainuma, who said the company has no current plans for deals. "We are so busy," he said in an interview. "It's more important for us to focus on the volume of work in front of us, rather than on M&A and pouring managerial resources into a new arena." In a time of both misinformation and too much information, quality journalism is more crucial than ever.

Nvidia is competing with OpenAI and Anthropic by acquiring the platforms that threaten its business model. Nvidia CEO Jensen Huang delivers a speech during the Computex 2026 exhibition in Taipei, Taiwan on June 1, 2026. Get your news from a source that's not owned and controlled by oligarchs. On Thursday, Nvidia, the tech giant that develops chips that power AI tools like ChatGPT, announced its acquisition of Hugging Face for nearly $13 billion amid an arms race to dominate every level of the industry. Well, specifically, the cost was $12,930,300,000, with 129,303 apparently being the character encoding for the emoji, which the startup Nvidia acquired is named after . Hugging Face is the venture that broke into AI safety discussions when OpenAI's advanced system breached its platform while undergoing a test .

Midjourney, the company best known as a maker of AI image and video generation tools, has decided to get into astrology. The company has announced it's purchasing Co-Star, a popular astrology app, and bringing on its CEO Banu Guler as its new Chief Design Officer. The app will still be under Guler's "complete control," according to the company, but now with Midjourney's resources behind it. In a post on X, Guler characterizes the acquisition as a natural extension of her friendship with Midjourney founder David Holz and their shared belief that "computers can be tools that reveal our humanity to ourselves. Tools that show you what you mean, how you feel, what you're imagining, before you can articulate it yourself."

Learning effective policies for adaptive data acquisition remains challenging: posterior-based methods rely on surrogate models and posterior approximations that can be misspecified or biased, while direct policy-learning methods map from historical observations and fail to exploit available model representations, making learning harder. We introduce policy learning with belief representations (POLAR), based on the insight that optimal data acquisition depends on the observation history only through a sufficient belief state. Specifically, POLAR decouples representation learning from policy learning by leveraging pretrained predictive foundation models as belief-state encoders, training a policy head on top of their representations. This yields a simple, unified amortised policy learning framework for Bayesian experimental design, Bayesian optimisation, and active learning, differing only in the task-specific utility used to train the policy. Empirically, we find that POLAR outperforms state-of-the-art amortised methods across diverse tasks while requiring far fewer training samples, demonstrating a significant step in the scalability and efficiency of amortised data acquisition.

Frameworks for ensuring fairness in machine learning typically focus on learning fair models from existing data. But this endeavor is often undermined by biases already present in that data. We therefore look to modify the data acquisition process itself to help gather fairer data that is inherently more suitable for training fair predictors. To this end, we introduce FairBED, which provides novel formulations for quantifying the fairness of datasets themselves based on the idea that fair datasets should be uninformative about sensitive attributes. We then use this to construct practical fairness-aware Bayesian experimental design (BED) objectives that maximize expected information gain about the target quantity of interest while minimizing expected information gain about sensitive attributes. We further derive a theoretical link between FairBED and demographic parity, and show empirically that models trained on data gathered using FairBED provide improved fairness-accuracy trade-offs compared to randomly acquired data and conventional BED.

Sequence schemas are abstract, reusable knowledge structures that facilitate rapid adaptation and generalization in novel sequential tasks. In both animals and humans, shaping is an efficient way for acquiring such schemas, particularly in complex sequential tasks. As a form of curriculum learning, shaping works by progressively advancing from simple subtasks to integrated full sequences, and ultimately enabling generalization across different task variations. Despite the importance of schemas in cognition and shaping in schema acquisition, the underlying neural dynamics at play remain poorly understood. To explore this, we train recurrent neural networks on an odor-sequence task using a shaping protocol inspired by well-established paradigms in experimental neuroscience.

KYOTO - Nidec President Mitsuya Kishida has said the major Japanese motor maker will suspend business acquisitions for the time being to focus its efforts on reconstructing the firm rocked by accounting and product quality fraud. Business acquisitions have been a growth driver for Nidec, based in Kyoto. "I will work on rebuilding our company's governance system," Kishida said in an interview Friday, showing a plan to spend ¥130 billion over five years on measures to prevent irregularities. A panel of outside experts that investigated the accounting fraud has concluded that excessive pressure from Nidec's founder, Shigenobu Nagamori, on company staff to meet performance targets was among the factors behind the irregularities. Pointing out that Nidec had "a corporate culture to pursue short-term profits," Kishida said, "We will build a system that makes it impossible to commit irregularities regarding accounting and product quality control." On future business management, he said, "We will review our operations, including the possibility of ceding what we have in our group to partner entities," suggesting that consolidating some of its existing operations could be an option.

Benchmarks increasingly guide deployment, procurement and scientific screening, yet a score supports only the response it records, not necessarily the deployment action. We introduce deployment-complete benchmarking, which tests whether benchmark evidence determines a deployment action. A benchmark is complete for a claim exactly when the action is constant on each evidence fiber; mixed fibers expose missing deployment information, and completion curves quantify the evidence required to resolve ambiguity. In controlled response spaces, benchmark-channel conformal coverage of 94.98% transferred poorly to an unmeasured deployment channel (10.07%), whereas response-rank intervals achieved 94.91% coverage; even zero benchmark error certified only 45.4% of candidates at the largest residual size. Public audits revealed incompleteness, including 97.9% mixed Tox21 fibers and zero median certifiable fraction in main Matbench and JARVIS audits. In held-out replays, certify-then-acquire reduced false decisions from 1.19% to 0.027% in Tox21 and from 20.3% to 0.128% in JARVIS, while changing model choice and identifying deployment-relevant probes. Deployment-ready benchmarks should report evidence, supported actions, ambiguity and completion cost rather than scores alone.

The validity of statistical inference depends critically on how data are collected. When data gathered through active data collection (ADC) are reused for a post-hoc inferential task, conventional inference can fail because the sampling is adaptively biased toward regions favored by the collection strategy. This issue is especially pronounced in black-box optimization, where sequential model-based optimization (SMBO) methods such as the tree-structured Parzen estimator (TPE) and Gaussian process upper confidence bound (GP-UCB) preferentially concentrate evaluations in promising regions. We study statistical inference on actively collected data when the inferential target is constructed in a data-dependent manner after data collection. To enable valid inference in this setting, we propose post-ADC inference, a framework that accounts for the biases arising from both the active data collection process and the subsequent data-driven target construction. Our method builds on selective inference and provides valid $p$-values and confidence intervals that correct for both sources of bias. The framework applies to a broad class of ADC processes by imposing only assumptions on the observation noise, without requiring any assumptions on the underlying black-box function or the surrogate model used by the SMBO algorithm. Empirical results also show that post-ADC inference provides valid inference for data collected by GP-UCB and TPE.