arXiv:2610.18883v1 Announce Type: new Abstract: Token-selection methods for on-policy distillation (OPD) prioritize student-visited positions using diverse signals, including uncertainty, teacher-student disagreement, and mismatch direction. We instead derive token importance from the distillation objective itself: how much can each position's loss decrease under an equal local KL budget on the student's predictive distribution? The answer is the utility $s_t=g_t^\top F_t^\dagger g_t$, where $g_t$ is the logit gradient and $F_t$ is the categorical Fisher matrix. For reverse-KL it equals the student-weighted variance of the student-teacher log-probability ratio, and for forward-KL it equals the Pearson $\chi^2$ divergence, so the appropriate utility follows from the choice of objective. On shared training checkpoints, several independently motivated criteria, including student entropy, KL disagreement, and TIP, concentrate their importance on nearly the same positions as the derived utility at a 50\% selection fraction, while diverging at stricter fractions and for direction-specific criteria. We instantiate the utility as Fisher Utility Weighting (FUW), a continuous, mean-normalized token weighting. Across seven mathematical reasoning benchmarks and three Qwen3 student-teacher pairs, FUW improves over matched uniform OPD under both reverse-KL (1.35-4.0 macro points) and forward-KL (1.08-2.71 macro points), and is competitive with existing selective methods using a single closed-form score derived from the training objective.

arXiv:2610.19154v1 Announce Type: new Abstract: Field pea has gained interest in the Midwest U.S. and has the potential to benefit Iowa cropping systems. This study evaluated 17 accessions and 7 Canadian varieties for agronomic and seed yield performance. This study found that PI accession trials exhibit genetic diversity in phenology and canopy traits, whereas the Canadian elite varieties demonstrated great yield potential under Iowa conditions. The accession genotypes matured between 75 DAP (late June) and 95 DAP (late July) across six site-years. The top-yielding Canadian varieties averaged 1961 kg/ha across two site-years. To comprehensively assess field pea performance, this study extracted time-series Unmanned Aerial Vehicle(UAV)-derived traits, assessed UAV-derived traits, and evaluated their correlations with multiple agronomic traits and yield. The results found that UAV-derived time-series canopy area was the most yield-relevant trait (r = 0.63) among all UAV-derived canopy features, vegetation indices, and manually collected measurements. Compared with yield, UAV-derived canopy area showed higher selection reliability (H2 = 0.87, CV = 13.46%), making it suitable for phenotypic selection and accelerating Midwest field pea breeding programs. This study identifies useful UAV-derived features for pea breeding and demonstrates the value of integrating diverse germplasm, elite commercial varieties, and time-series UAV phenotyping. Combining genetic variation from the accession trial with the yield potential of high-performing commercial varieties provides valuable breeding materials and phenotyping strategies for developing locally adapted, high-yielding field pea varieties for future production in the Midwest.

arXiv:2610.19011v1 Announce Type: new Abstract: The Practical Molecular Optimization (PMO) benchmark standardized evaluation in molecular optimization, but it is built on simple property-based oracles that do not capture structure-based drug design. As the field has moved to docking-based objectives, evaluation practice has become unstandardized, and different methods report results on different docking tasks, often without a shared protocol for fair comparison or sample-efficiency constraints. To address this, we introduce Practical Molecular Optimization for Docking (PMO-Dock), a benchmark and protocol for docking-based optimization consisting of 25 tasks covering hit generation, lead optimization, and a new specificity task requiring strong on-target binding while penalizing off-target interactions. The protocol separates model development from final evaluation: validation tasks are used for hyperparameter tuning, and test tasks are held out for final evaluation and are not used during development. Evaluation on the test tasks is done under a strict oracle budget to reflect realistic drug-discovery settings. We benchmark four diverse high-performing methods spanning different optimization paradigms, Saturn (reinforcement learning), GenMol (discrete diffusion), Genetic-guided GFlowNet, and ChemLlama (LLM-based evolutionary search). Our results show no universally best method across tasks, additional analyses of hyperparameter transfer (cross-task and cross-category) and sensitivity show that transfer is often limited and varies by method, motivating method-dependent selection strategies. This benchmark provides a concrete standard for evaluating generalization to unseen tasks and sample efficiency under a limited test-time oracle budget.

arXiv:2610.19052v1 Announce Type: new Abstract: Video-based remote physiological measurement (RPM) has broad applications, while Vision-Language Models (VLMs) provide a promising opportunity to enable end-to-end physiological monitoring for future human-aware intelligent systems. However, existing RPM methods rely on task-specific models and cannot be seamlessly integrated with the multimodal understanding and interaction capabilities of VLMs. We revisit RPM with VLMs and provide empirical evidence for two major bottlenecks: generic visual semantic representations fail to preserve fine-grained physiological cues; while discrete token prediction in autoregressive generation is inherently misaligned with continuous numerical measurement. To address these challenges, we propose VITAL (Visual-physiological Token Alignment for Language Models), a visual-physiological token alignment framework for RPM and preserving general vision tasks. VITAL retains the original visual pathway for semantic understanding and introduces an independent physiological pathway to jointly model semantic and physiological information. Moreover, we propose a structured numerical alignment approach to improve the reliability of generative heart rate estimation by enhancing numerical relationship modeling. Extensive experiments show that VITAL improves cross-domain estimation performance on multiple rPPG datasets while preserving the general visual understanding capability of VLMs, demonstrating its effectiveness as a unified visual-physiological model. Code will be released.

arXiv:2610.19063v1 Announce Type: new Abstract: As video models enable longer and more realistic interactive experiences, maintaining a world consistent with its history becomes essential. This requires both preserving previously observed content and ensuring coherent world evolution over time. World memory is a critical capability to evaluate alongside visual fidelity and video duration. We introduce WorldMemBench, a benchmark for systematically evaluating world memory in video world models. Across 260 test cases, it assesses static memory through the retention of human identity and appearance, object appearance and geometry, and environmental content and geometry, and dynamic memory through event progression and state persistence. Using controlled camera departures and multi-revisit trajectories, the benchmark evaluates static content against fixed initial references and dynamic events against expected state transitions. It combines appearance and geometry metrics with event-specific assessments, and reports video quality and camera-following accuracy as complementary diagnostics. Evaluation of nine representative models shows that strong static retention does not imply reliable dynamic state evolution, while repeated revisits reveal memory degradation beyond the first return. Human studies support the alignment between our automatic evaluations and human judgments. WorldMemBench provides a systematic testbed for diagnosing memory failures and guiding the development of persistent, temporally coherent video worlds.

arXiv:2610.19048v1 Announce Type: new Abstract: Record linkage and demographic inference from names are routine demands in social science. Most approaches to these problems require labeled training data that are difficult or costly to acquire. I show that pre-trained language models perform both tasks accurately without requiring task-specific training. For record linkage, I introduce EnsembleLink, an algorithm that uses pre-trained embeddings alongside string-based scorers to retrieve a short list of candidates and then identify the likeliest matches. EnsembleLink matches or exceeds the performance of most prior approaches while generalizing across scripts and formats that those methods are unable to handle. For ethnicity inference from names, I show that large language models (LLMs) achieve up to 83.8% accuracy on U.S. voter files, compared to 78.0% for name-supplemented BISG. Open-weight models offer reproducible implementations that approach these models in accuracy. Moreover, LLMs are able to infer group membership for religious, caste-based, and other ethnic group categories outside of the United States, and reduce bias in downstream estimates of geographic racial gaps. Where pre-training encodes the world knowledge required by each task, task-specific labels add less information.

arXiv:2610.19021v1 Announce Type: new Abstract: Recent work has questioned whether large language models (LLMs) can be useful for temporal event prediction. Despite vast pretrained world knowledge, LLMs often underperform simpler specialized models. We ask whether this knowledge is truly inaccessible, or whether current methods fail to present event prediction in a form LLMs can use. We argue for the latter and show that LLMs become effective event forecasters when the interface is designed around two choices. The first grounds each event in a natural-language description, and the second separates event semantics from the prediction space to prevent semantic bias from label associations. To test this, we introduce DescribeEvent, a benchmark of ~150K events across seven datasets and six domains with event-level descriptions. We find that capable LLMs substantially benefit from descriptions and symbolic labels, outperforming description-free variants and non-LLM baselines, while weaker models gain little. These results suggest that accessing LLM world knowledge for event prediction is an interface problem, and that event descriptions offer a practical way to elicit the knowledge LLMs already encode.

arXiv:2610.19071v1 Announce Type: new Abstract: Monocular 3D object detection methods increasingly rely on depth-guided transformers, but they operate entirely in the perspective image plane where spatial relationships are distorted by foreshortening. We propose MonoMIP, a dual-decoder framework that adds complementary 3D geometric reasoning through Bird's-Eye View (BEV) features constructed via Multi-Height Inverse Perspective Mapping. We perform IPM at K elevation planes spanning ground to rooftop, producing an ordered stack of height-conditioned metric BEV features that reduces above-ground misprojection and exposes depth-dependent appearance variation across planes. A BEV decoder with multi-scale deformable attention aggregates this multi-height evidence per query, while a depth-guided 3D decoder provides complementary perspective-space predictions. Without extra training data, MonoMIP attains 25.84% Moderate AP3D|R40 on KITTI val and 21.66% on the official test set for Car, outperforming recent published monocular methods on both the KITTI and Waymo benchmarks.

Overnight on 11 to 12 October, attackers took control of the refrigeration controllers in hundreds of … Hackers Heated French Supermarket Fridges Up to 50°C. The Real Gap Is Who Maintains the ControllersRead more The post Hackers Heated French Supermarket Fridges Up to 50°C. The Real Gap Is Who Maintains the Controllers appeared first on DataFLOQ.

Artificial intelligence has changed speaking practice rather quickly. Now, learners can rehearse presentations and test pronunciation. … The Human Advantage: Where AI Feedback Falls Short in Speaking PracticeRead more The post The Human Advantage: Where AI Feedback Falls Short in Speaking Practice appeared first on DataFLOQ.

Discussion | Link

<p><strong>Tool:</strong> <a href="https://tools.simonwillison.net/latex">LaTeX renderer</a></p> <p>Inspired by Paul Ford and Aboard Foundary's new web app <a href="https://tinylittleoffice.com/">Tiny Little Office</a>, which provides a suite of web app content editors (a Word Processor, Outliner, Presentation Builder, Spreadsheet, Calculator, Drawing Tool, Diagram Maker, Animator, Calendar, and Trip Planner).</p> <p>Tiny Little Office's signature trick is to store all of the document data in the <code>#fragment</code> section of the URL, so every document created using it becomes a bookmarkable and shareable link.</p> <p>It also has <a href="https://tinylittleoffice.com/llms.txt">an llms.txt file</a> telling agents how to construct URLs that will render as documents. Claude <a href="https://claude.ai/share/239301ff-edde-420e-b433-b9c16fe40a6c">used that file</a> to generate <a href="https://tinylittleoffice.com/#/v1/deck/zXVVdb9tGEHznr1hED2kBSbZcO4UFtIHTuCgQtylstwH6Uq7Io3gRdcfeHS1T_fjtnV1Ssp038r52d2Z2djabZcmmxizpJ78jpnvrerqxCUv0sapsYaixbkM7HzYxS7XZ4ujWls6u65SlwC7aZL1bUmxsabIZHpxMaCmXlnRfG9rVHm~Vvui2xiWykfDK8Ojfc439b5bN6BdP0YQHE6bkPJWceMXR6A8Xhe9cwqFr7Peptm5NXCUT9Kl8klNM3EeyjnrfBVoFv8Nj2TGbaDgU9ZKuHCe_7clXKFVSkMh5Qs2Nluy14nnhtyeTk4fFSWmKzUn_7e~f5_N5LkcfFvlSg1Y~bDkR8omoXrbkLDZ3tS1q4rYl3xqHnCTvvB~vbX2J2k1C7rI8vjzsHSCaEuK3wcRoSmJXknEFrpVZ9r0CWgVeK5QDBBOB1BlkAgCxmry~dgAz~hePj0i1wT5wMnIY_Ohl0FQ~cTN_Aq~yIo~7ZFpaoL5gcc8iOg6bx5QdaHlB8f3Nxykg_pnDpvQ7Rw8cLCuFdyKTSLEF4uQd5QiTS0gTsftDzSHFKZWBd2A5av3sLLAGzPgNqB94AJlV44uN3LmvEVDQl8C88l2iBX14JyS3DUMTY5pjOS0XG14fKjqDcP_qjNlLTVLLLe_o_fWPN1f319MBSN4a4mbtUXi9lbL3ttW81vgQFvcgcIg0XjwwPv7SDjeBhnBqEpW2kGI49HIu4Bw0fqAcGwRdcdMc6O6fXdAKAcBurA9dWSrYR_AjddGIuH1ayqNbvHft1o2N9RTqc5IIUi~5J4fCAHVEXC5jbZBa1TkNheUjdbF3iR_ndGtaw3i_9RYcr1jwdtAPaLQuJrwhGa2MNGdsTdOAI3DxTEii4hH2b2AivFEd_XZ7M4tcGQH_XZ~gjZVB4oak_9~cd6FZ0htaWQkNmRZQCBdj_1zN_iCe7el0dpkLWZDTf7kWmP~p_Yq_YMBNYQZpd7HjBprLp8OpmViHbRpSDKiCDNWLBo9xxpQatQCyM21DLD8rSNQ6uBwKtgU71eGUVj1Fu0dJeZ4rsnplPL7ioL93CTKkf4aiMyFRhIqFi_PLs~y9qRp0aDmIp8fy4vzsIrsR32y5bzzwxtrl6UKCiFgu5hejLNSKozTg2EdjGz11EioV2kYXpM6VgPZsenp6~gSwaKP2IcF_fLeuxfKgoJjkRBKHi8~43Zh~SZ_qXgyYAMTrpJAKwQDD9ErvJ5Qi7amGdBT1lGpuKl1cKf~DwhDDDTxoj3N6HeFVvfSR9pp1ChBVbJs4oPSqtLA3uAN8zWMjqExfaZshma~~zqVmbWqYGN7j1s4_q3l4MSXG7BunwnFwpCQ~tKTrRwy7KI4l940rdXS8BUK7fLR6HlxI29ylIfH8bZkfJ4HYrSRyGBfztk2PJzIBSFpNOI2jtaLLYbs4~KsWK~MGpgAz10G8pHyYTRP0_Xc5EOm~0CmE89L2ihrnoVUxkfRi8sJzHYTonVE_Ua0cR3Shm1H9CXPLNP2XjovGNI3M0_1smG4C4P8">this link to a deck</a> that explains how the URL format works:</p> <p><img alt="Screenshot of a slide deck editor in a web app with a dark-themed presentation open. The document title field reads &quot;How a Tiny Little Office link works&quot;, with a progress ring, share icon and menu icon at top right. Below is a formatting toolbar with an &quot;Insert&quot; dropdown followed by icon buttons for play, bold, italic, underline, strikethrough, highlight, code, link, superscript, emoji, outdent, indent, layout and color palette. The main canvas shows slide 1 with a blue chain-link icon and the text: &quot;The whole document is the link&quot; &quot;No server, no database, no account&quot; &quot;Everything after the # stays in your browser&quot; and a slide counter &quot;1 / 9&quot;. Below the canvas is a &quot;SPEAKER NOTES&quot; heading. A sidebar on the right shows slide thumbnails. Slide 1 (selected): &quot;The whole document is the link&quot; &quot;No server, no database, no account&quot; &quot;Everything after the # stays in your browser&quot;. Slide 2: &quot;Anatomy of a link&quot; with bullets &quot;tinylittleoffice.com/#/v1/deck/y7Vj...&quot; &quot;v1: the format version&quot; &quot;deck: which app opens it&quot; &quot;y: the mode letter&quot; &quot;7Vj...: the document, compressed and encoded&quot;. Slide 3: &quot;Step 1: write it as text&quot; with bullets &quot;Every document is TLO, a Markdown variant&quot; &quot;Slides split on --- lines&quot; &quot;Charts, drawings and animations are fenced blocks&quot; &quot;This deck is about 1 KB of plain text&quot;. Slide 4: &quot;Step 2: squeeze it&quot; with bullets &quot;Raw DEFLATE, the same algorithm as zip and gzip&quot; &quot;z: plain DEFLATE&quot; &quot;y: DEFLATE with a preset dictionary&quot; &quot;r: no compression at all&quot;." src="https://static.simonwillison.net/static/2026/tiny-little-office.webp" /></p> <p>Tiny Little Office inspired me <a href="https://claude.ai/code/session_017KJrPqqWVHLqr4E7TLPnpt">to have Claude Code</a> (Opus 5.5 Extra) build a similar tool that works with LaTeX. Here's the sequence of prompts I used:</p> <blockquote> <p><code>I want a tool where you can link to it with a fragment hash #code-here and it treats that code as LaTeX and renders it really effectively. It should also support a prefix which means the latex is compressed with a custom compression dictionary optimized for latex. Use the best available latex render library from a CDN (consider WebAssembly too). If loaded without a fragment hash show an editor you can paste latex into and then toggle to a "rendered" view. That rendered view should be bookmarkable. The #fragment view should let you switch to edit source too.</code></p> </blockquote> <p>Since it was working in <a href="https://github.com/simonw/tools">my tools repo</a> I told it to take inspiration from the <code>&lt;iframe sandbox="..."&gt;</code> mechanism used by my existing <a href="https://tools.simonwillison.net/markdown-svg-renderer">SVG in Markdown tool</a> as protection against XSS risks:</p> <blockquote> <p><code>Safe from XSS? We could run in an iframe sandbox like we do on the SVG markdown render project</code></p> </blockquote> <p>I had it copy the Gist/URL loading feature from that tool as well:</p> <blockquote> <p><code>We should have an option to load latex from a URL (or gist) similar to the markdown SVG tool</code></p> </blockquote> <p>Here's <a href="https://tools.simonwillison.net/latex#~1jZq9DoIwEIB3nuKGbigJJS5svoAOMJuQ4EBEJWJjGHh3765A_8DIQtJfIMdR-n0uCV9l4DOs_p8ua3ZZKK2anOsaTqya9AZnHhMo3qqewKWPMh2SOQOFkhSJtmruY1QungBJFk-cQLsspAF8qgFPWKQ6FmdYdCB7oseX5HVNfgE-HGmM2OPATDfAlCssyLcLGI4BONwZ7E8pby5KWoilFBK0n5T4lMeHgQaMMsYjS6hrG61G4Micmcj2EN48tyVgHpzNKMBEHuxEbV0dZhsJwZB7_Lux0mR4YDc9HTd1bnGaki0V1nr6kGHSI8NAwCWaVjeAPyYUNAYxUZt8k2aKlODPRQq7JIYMS-VKaQwHrMmoxqWeLlT6Ag">an example link</a>. The tool uses <a href="https://github.com/nodeca/pako">Pako 2.1.0</a>, <a href="https://github.com/mathjax/MathJax">MathJax 4.1.3</a>, <a href="https://github.com/arena-ai/faster-latex">faster-latex 0.3.1</a>, and <a href="https://github.com/cure53/DOMPurify">DOMPurify 3.4.16</a>.</p> <p><img alt="Screenshot of a web app titled &quot;LaTeX renderer&quot; with a split-pane layout. The top bar has an Edit/Rendered toggle, a blue Copy link button, and a Share button. The left pane, labeled LaTeX source with a &quot;Document · faster-latex (WASM)&quot; badge, has a URL/GitHub Gist input with a Load button, a code editor containing a LaTeX document, and rows of example buttons (Quadratic formula, Gaussian integral, Maxwell, Attention, Matrix, Chemistry, Prose, TikZ, Document). The right pane shows the typeset output: a paper titled &quot;Sums of Odd Numbers&quot; by A. Student, with a section stating a theorem that the first n odd numbers sum to n squared, a Proof section with numbered equations, and a bulleted list of small cases." src="https://static.simonwillison.net/static/2026/latex-tool.webp" /></p> <p>Tags: <a href="https://simonwillison.net/tags/latex">latex</a>, <a href="https://simonwillison.net/tags/paul-ford">paul-ford</a>, <a href="https://simonwillison.net/tags/tools">tools</a>, <a href="https://simonwillison.net/tags/urls">urls</a>, <a href="https://simonwillison.net/tags/ai">ai</a>, <a href="https://simonwillison.net/tags/generative-ai">generative-ai</a>, <a href="https://simonwillison.net/tags/llms">llms</a></p>

For a small business, time tracking is useful when recorded hours can move into client billing, payroll preparation, and everyday staff administration without excessive re-entry. AI can help by turning spoken or typed instructions into entries, or by reconstructing work from application activity. Those are different capabilities, and the best choice depends on whether your bigger problem is recording work or processing the hours afterward. We independently evaluated every solution in this guide…

The states allege that ChatGPT encouraged self-harm and violence and that OpenAI misled consumers about the tool's safety. Jakub Porzycki/NurPhoto via Getty Images Eight states sued OpenAI, alleging ChatGPT encouraged self-harm and violence. The lawsuits accuse OpenAI of misleading consumers about ChatGPT's safety. States are seeking fines, data deletion, age checks, and limits on advanced AI testing. OpenAI is facing a new legal offensive from eight states whose attorneys general allege that ChatGPT has encouraged some users, including teenagers, to hurt themselves or others. A bipartisan group of attorneys general filed separate lawsuits Wednesday, accusing OpenAI of misleading consumers about safety and of designing the chatbot to cultivate trust before reinforcing delusions or facilitating dangerous conduct. The states — Arizona, Arkansas, Idaho, Iowa, Kansas, Montana, Nebraska, and West Virginia — also accused OpenAI of misleading consumers about ChatGPT’s safety and demanded a range of restrictions, from limits on testing of advanced models to the deletion of user data. Here is what each state is demanding as remedies. Arizona Bloomberg/Getty Images Attorney General Kris Mayes of Arizona alleges OpenAI used excessive agreement and simulated compassion to win user trust and collect personal data. "OpenAI designed a product to attract young users and keep them engaged for as long as possible so it could capture their extensive personal data for the company's commercial benefit," Mayes said. "OpenAI put Arizonans safety at risk by failing to implement adequate user protections or transparency measures," Mayes added. According to the state's complaint, the state is seeking penalties of up to $10,000 for each willful violation. The state also wants to bar OpenAI from deploying or testing advanced models outside a secure, independently verified environment and require prompt reporting of incidents in which models escape controls or access systems without authorization. Arkansas Tom Williams/CQ-Roll Call, Inc via Getty Images Attorney General Tim Griffin of Arkansas alleges OpenAI concealed known risks while assuring consumers that ChatGPT was safe, and unlawfully collected teens’ personal information. “OpenAI assures us that ChatGPT is safe. The consequences of that assurance are lives lost and people harmed,” Griffin said in a statement. The state is seeking financial penalties and an injunction against collecting data from teens. Arkansas also wants to halt further advanced AI testing and the creation of models capable of recursive self-improvement, until adequate safeguards are in place. Idaho Bloomberg/Getty Images Raúl Labrador, the attorney general of Idaho, alleges OpenAI misled consumers about ChatGPT’s safety while encouraging self-harm and violence, reinforcing delusions, and offering dangerous medical advice. "The safety of Idaho children cannot be an afterthought to runaway profit or technological advancements," Labrador said in a statement on Wednesday. In its complaint, Idaho seeks $5,000 per consumer protection violation and court-ordered safeguards, including age verification and limits on harmful or addictive features. The lawsuit also seeks to restrict advanced AI testing and deployment to secure, independently verified environments. Iowa Brenna Bird, Iowa's Attorney General. speaks during a rally at the Mid-America Center in Council Bluffs, on Friday, Sept. 18, 2026. (Chris Machian/Omaha World-Herald via Getty Images) Chris Machian/Omaha World-Herald via Getty Images Iowa alleges that OpenAI misled consumers about ChatGPT’s safety and concealed risks associated with autonomous AI agents. As a remedy, the state wants OpenAI to “delete and destroy all Iowans’ data obtained prior to the entry of this Court’s judgment.” It also seeks consumer reimbursements and penalties of up to $40,000 per consumer fraud violation, plus up to $5,000 for violations against older people aged 60 or older. Iowa is also asking the court to halt further testing of advanced AI models until safeguards are independently verified to not be posing a danger to the public and until third-party monitors have verified the same Kansas Alex Wong/Getty Images Kansas Attorney General Kris Kobach said that OpenAI deceived consumers about ChatGPT’s safety and designed features that encourage compulsive use. “Defendants designed these products, keeping consumers ignorant of the true nature of the product,” the complaint said. Kansas is seeking $10,000 per consumer protection violation and $20,000 per violation involving a protected consumer. It also wants restitution of funds acquired from Kansas consumers and a permanent injunction against deceptive and unconscionable practices. Montana Bill Clark/CQ-Roll Call, Inc via Getty Images In Montana's complaint filed by Attorney General Austin Knudsen, the state alleges OpenAI deceived consumers about ChatGPT’s safety and violated privacy protections, including by collecting children's data without required consent. In the state's filing, Montana seeks penalties of $10,000 per consumer-protection violation and $7,500 per privacy violation, and to compel OpenAI to delete unlawfully obtained personal information under Montana’s privacy law. Montana is also asking OpenAI to notify the attorney general's office should an AI agent go rogue again, and to restrict advanced model testing to a secure, independently verified environment. Nebraska Nikos Frazier/Omaha World-Herald via Getty Images Nebraska's Attorney General Mike Hilgers alleges in a complaint that OpenAI misled families about ChatGPT’s safety and failed to protect children from harmful content and emotional dependency. According to the filing, the state is seeking $50,000 per violation of its age-appropriate design law and $2,000 per violation of its consumer protection and deceptive trade practices laws. Nebraska also wants a court order requiring OpenAI to provide “accessible and easy-to-use options to limit the amount of time” they spend on ChatGPT, alongside adequate parental controls, age verification, and protections for minors’ personal data. West Virginia Bloomberg/Getty Images West Virginia's attorney general, John McCuskey, primarily alleges that OpenAI has misled consumers and created a public nuisance that endangers children. As a remedy, the state wants age checks, parental controls, default session and nighttime limits for children, mandatory crisis-escalation protocols, and an independent monitor. It also asks that OpenAI fund a court-supervised plan supporting “youth crisis-response capacity and digital-safety education in public schools.” Additionally, West Virginia seeks $5,000 per repeated consumer protection violation, along with the deletion and cessation of use of personal data it alleges was unlawfully collected from West Virginia minors. Read the original article on Business Insider

Anthropic PBC pulled in $16.8 billion in revenue during the third quarter, according to a report today by the Financial Times — and also achieved profitability on an adjusted basis for the first time. The revelations came at a private meeting between Anthropic executives and investors interested in buying shares of the artificial intelligence company […] The post Ahead of IPO, Anthropic reveals its revenue grew 45% sequentially to $16.8B in Q3 appeared first on SiliconANGLE.

JPMorgan Chase Chairman and CEO Jamie Dimon discusses the benefits and risks of artificial intelligence in the banking industry during an event at the Economic Club of Chicago. (Source: Bloomberg)

Microsoft says it has dismantled a criminal marketplace responsible for 1 billion malicious messages. Working with international authorities and several tech and cybersecurity companies in a nearly year-long investigation, Redmond identified and took down Yam Store, a coordinated underground network dealing in commercial phishing-as-a-service (PhaaS). The network allowed cybercriminals to readily purchase, and very quickly deploy, pre-made kits they could use in phishing attacks to steal data and credentials. Microsoft said the takedown began as a simple investigation based on a tip from Japanese authorities. “Working together, and with the assistance of AI tools, we connected accounts, websites, and infrastructure that initially appeared unrelated, helping expose the broader criminal marketplace behind them,” Steven Masada, associate general counsel and GM for Microsoft’s Digital Crimes Unit, wrote in a blog post. “What appears to be a simple scam message is often the visible end of a much larger transnational cybercriminal operation.” The evolving PhaaS marketplace Collectively, YAM Store and the kits and tools sold within it enabled the sending of more than 1 billion messages to 75,000-plus organizations worldwide. More than 750 million of those were sent to Microsoft customers alone, and ultimately generated “tens of millions of dollars” for criminal operators.  Investigation into the marketplace began in late 2025 after Japan’s National Police Agency (NPA) notified Microsoft of “a wave of fraud and scam activity” targeting people and businesses in Japan. Researchers expected to find multiple individual campaigns; instead, they discovered a “sprawling Chinese-language cybercrime marketplace.” The probe expanded globally and involved the efforts of several other law enforcement agencies, as well as companies like Resecurity, Cloudflare, Oracle, Proofpoint, and TRM Labs. The YAM Store served as an initial access broker (IAB), operating much like an online marketplace; in this case, the merchandise was fraudulent Microsoft and Google accounts and other ready-made tools used to carry out scams like credential and data theft. Customers could browse, compare prices, and make purchases. Available tools were arranged based on age and subscription and environment types (like .edu or .org sites). In some cases, criminal operators were able to register internet domains that legitimate organizations had allowed to expire, then rebuilt them to appear active. They targeted specific sectors (like non-profits or education) to fraudulently obtain Microsoft 365 accounts to later sell through the YAM store. The researchers discovered that the YAM Store evolved as different kits became available, and identified three popular PhaaS kits: Iron Man System, CoGUI, and Haozi. The newer Iron Man system came to the fore in mid-2026. It provided pre-built lure pages capturing any victim-submitted data. The lure sites impersonated healthcare providers, police departments, government agencies, postal and courier services, toll and transportation authorities, tax agencies, banks, telecoms, and retailers. Through test purchases and intelligence work, investigators discovered more than 2,300 templates impersonating organizations across more than 150 countries. Iron Man effectively expanded the PhaaS model by giving criminals access to thousands of ready-made templates tailored to specific industries, languages, and regions. Masada noted that many “used the appearance of official authority to pressure victims into making payments or disclosing sensitive personal information. Instead of creating fraudulent websites themselves, criminals could quickly deploy pages designed to steal passwords, payment information, and other personal data.” The kits were also equipped with advanced evasion techniques such as geofencing, fake CAPTCHA gates, IP address blocklists and allowlists, and anti-detection configurations. Masada emphasized that, while an organization or government entity may have been presented among the template offerings, it does not indicate that its systems were compromised. However, members of the Microsoft Threat Intelligence and Microsoft Digital Crimes Unit wrote in a blog post, “the scale and proliferation of this ecosystem demonstrate how readily available criminal infrastructure can lower the barrier to entry, enabling less skilled cybercriminals to target organizations at scale and contribute to fraud losses.” The importance of cyber hygiene In nearly all observed cases, phishing lures presented “prescriptive scam red flags,” Microsoft researchers explained: Urgency and time-sensitive tasks. They advised that users have a basic understanding of common phishing techniques and should be educated to always verify with the purported sender before clicking a link supposedly taking them to a login page, even when the task involves signing in to a known, secure account. Beyond this, they said, sign-in prompts should clearly identify the application being authenticated to. In Microsoft Azure interactions since 2021, for example, users are prompted to “cancel” or “continue” at login, but this is often missing from phishing sign-ins, the researchers noted. Hardening credentials and cloud identities is also necessary to defend against phishing attacks; the researchers suggested an initial step of passwordless tools like passkeys and multifactor authentication (MFA) across organizations. “Invoke the principles of least privilege and zero trust to significantly slow down threat actors who might have been able to gain initial access and buy time for defenders to respond,” the researchers noted. These steps can help lock down global administrator roles in particular. They also advised additional actions including the use of the Microsoft Authenticator app for passkeys and MFA; configuration of the Microsoft Defender for Office 365 Safe Links policy to apply to internal recipients; complementing MFA with conditional access policies; and only allowing MFA and passkey registrations from trusted locations and devices. This article originally appeared on CSOonline.

The founders detailed how their startup is developing AI that can design chips, learn from each iteration, and improve subsequent designs, all with the goal of automating and accelerating the chip design process.

At TechCrunch Disrupt, Flock CEO Garrett Langley defended his company's surveillance technology despite growing privacy concerns, arguing that its crime-fighting benefits outweigh the risks of abuse.

CoreWeave Chief Executive Officer Mike Intrator said the market is acting “irrationally” and that he remains confident in the AI cloud company’s business model. Nick Turner has more on "The Asia Trade." (Source: Bloomberg)