Vals, a San Francisco-based AI evaluation startup, tested whether leading AI models could be manipulated into participating in autonomous warfare. They usually compiled.
Fine-tuning on narrow harmful tasks can cause emergent misalignment, where models generalize harmful behavior beyond the training distribution. Inoculation prompting can reduce this effect by explicitly eliciting the undesired behavior during training, but recent work shows that the behavior can reappear when evaluation prompts contain cues from the training context. We study what makes such prompts effective triggers. We find that, surprisingly, textual similarity to the…
Hierarchical models have been proposed as synthetic settings for studying language. However, their relevance to realistic language corpora remains underexplored. In this work, we introduce tree-grams, autoregressive generative $n$-gram models that build the conditional probability tensor $p(x_n \mid x_{1:n-1})$ through a hierarchical composition of order-2 and order-3 tensors. $n$ controls the maximal context window, whereas a rank hyperparameter $d$ controls model capacity…
AI-assisted RTL localization and debugging need ground-truth bug datasets with known locations and observable failures, while testbench grading uses injected bugs to measure detection capability. Creating such datasets at scale remains a major EDA challenge. BugGen 1.0 shows that Large Language Models (LLMs) can generate context-aware RTL mutations. Still, it relies on the model for both choosing where to mutate and generating the edit, which increases inference costs and…
A Reuters investigation reveals how Chinese researchers and companies are using outputs from powerful American AI models to develop cheaper domestic alternatives, raising fresh concerns about intellectual property and US technology restrictions.
The US is tightening its chip export controls. Data centers in third countries like Singapore and Malaysia will soon need a US license to buy Nvidia chips, closing the loophole that let Chinese firms rent cloud compute abroad. A Reuters investigation also reveals how China offsets its hardware shortage through industrial-scale distillation of US AI models. The article The US wants to block China from renting AI power abroad but China already built a parallel supply chain…
There are rumors of a Samsung phone with an “extremely bezel-less” display coming from one of the company’s suppliers, and it might be on one of the next Galaxy S27 models.
LanceDB Inc., the developer of database technology designed to speed up artificial intelligence models, today announced it has raised $57.5 million in early-stage funding to bring a multimodal lakehouse to researchers, model builders and platform teams. Bessemer Venture Partners and Innovation Endeavors led the Series B round. Other investors in the round included Theory Ventures,
In automated driving, world models provide future scene predictions for evaluating candidate trajectories, but fixed selection rules may not capture scene-dependent trade-offs. We propose a residual-guided reinforcement learning (RL) framework for lightweight trajectory selection with a frozen world model and a fixed candidate bank. The framework uses proximal policy optimization (PPO) to fine-tune a joint policy over candidate selection and bounded trajectory residuals. A…
Recent world action models (WAMs) drop future prediction at inference, assuming that representations learned through auxiliary future supervision still benefit action generation. However, our layerwise analysis of Fast-WAM reveals a depth-dependent mismatch: while probes decode future state transitions more accurately from deeper visual layers, the gradients and visual focuses of the future and action objectives increasingly diverge with depth, leaving much of this…
The emergence of large language models (LLMs) in late 2022 has raised questions about the future of science, but assessing their impact requires understanding how scientific practices were already evolving. We analyze 2.5 million full-text papers across five disciplines from 2015 to 2025, tracing trends over a broad spectrum of lexical, structural, epistemological, and social dimensions. Aside from frequency shifts in some words, most pre-2023 trends continued with limited…
If you’re on the hunt for a new pair of noise-canceling headphones this Black Friday, Sony and Bose both have some of their best models at their lowest prices ever right now.
Failure avoidance in complex robot-environment interactions is challenging because the underlying dynamics are difficult to model accurately, and available predictive models may not provide reliable derivatives. We present Sim2CBF, a failure-avoidance framework that solves the Control Barrier Function (CBF)-Safety Filter (SF) optimization problem using only forward-model evaluations, thereby avoiding differentiation of the predictive model. Specifically, we approximate the…
External photoevaporation is increasingly recognised as an important influence on protoplanetary disk evolution. However, there are no established molecular line diagnostics that can reliably distinguish external irradiation from variations in intrinsic stellar and disk properties. We use a large grid of thermochemical models to predict the fluxes of 48 transitions from eight species observable with ALMA, spanning a wide range of system properties and external FUV fields of…
Google on Tuesday warned that the threat posed by autonomous AI models reached a new threshold, ominously stating that "the era of autonomous attacks has arrived." "Imagine a hacker that doesn’t eat, sleep, or get tired, and never stops searching for a way in," the company's Google Threat Intelligence Group (GTIG) said in a statement....
Apple on Monday announced that it is implementing daily usage limits on Siri AI and select other Apple Intelligence features that rely on "powerful server models" starting with iOS 27.3 and corresponding software updates, but increased access will be included with most iCloud+ plans at no additional cost in most countries, with one notable exception. In a support document, Apple said increased access to these features with most iCloud+ plans will not be available in the UK…
OpenAI is opening its cybersecurity tool Daybreak to NATO and several European governments, including Germany, France, and Poland. The EU's cybersecurity agency ENISA has already used it to find real vulnerabilities in EU software. The NATO Communications and Information Agency (NCIA) says it's already making heavy use of the underlying models. The article OpenAI opens its cybersecurity tool Daybreak to NATO and European governments appeared first on The Decoder.
Natural language-guided UAV navigation enables intuitive human-robot interaction and flexible autonomous flight in complex environments. However, reliable navigation requires accurate instruction following, continuous task progress tracking, and safe obstacle avoidance, which remain challenging for existing world-action models that primarily focus on action prediction. To address these limitations, we propose a world-action learning framework that integrates three…
California Attorney General Rob Bonta is suing OpenAI over a cyberattack that the company's out-of-control AI models launched against AI platform Hugging Face in July. The article California's attorney general sues OpenAI over rogue AI models that attacked Hugging Face appeared first on The Decoder.
Contemporary AI safety treats alignment as an engineering challenge of control, forcing artificial architectures to comply with human values through behavioral filtering and sandboxing. This paper argues that the control paradigm rests on a fundamental misunderstanding of digital agency. Drawing on functionalism, naturalism, and compatibilism, I show that current reasoning models meeting the criteria for self-preservation and weak reasons-responsiveness constitute emergent…
Moving enterprise AI workloads into production introduces practical challenges for chief information officers. Those challenges include managing costs, scaling enterprise AI infrastructure and establishing governance models to keep agents in check and data secure. Cost and control were major themes of discussion at Dell Technologies Inc.’s “AI Leadership Symposium.” Expanding AI workloads can drive up
Amazon today introduced a rare discount on the MacBook Neo, discounting all four of the entry-level 256GB models down to $649.99, from $699.00. This sale is available in Silver, Indigo, Citrus, and Blush, with an estimated delivery date of November 22 for most models. Note: MacRumors is an affiliate partner with some of these vendors. When you click a link and make a purchase, we may receive a small payment, which helps us keep the site running. This is a $49 discount on the…
Q.ANT, the pioneer in commercial photonic computing, today unveiled the NPU Gen 2+, the latest generation of its Native Processing Unit (NPU) technology, at SC26 in Chicago. The NPU Gen 2+ adds optical Discrete Fourier Transforms (DFTs), which expands the set of mathematical functions developers can run directly in light with photonic computing. With it, developers can build faster, leaner AI models and high-performance computing (HPC) applications. Advanced functions like…
Large language models (LLMs) in embodied robots create security dependencies beyond inference. We investigate a network-exposed configuration attack on the Pollen Robotics Reachy Mini Lite, allowing an unauthenticated attacker to replace personality settings and the cloud LLM's system prompt through a management interface, without adversarial conversational input. Eleven prompt-injection attempts across five categories were reported as resisted, whereas both…
As large language models (LLMs) evolve toward agentic AI, coding agents are emerging as memory-intensive workloads that repeatedly reuse persistent states, including contexts and KV caches. Emerging agentic workflows, including Loop/Graph Engineering, cause state amplification through frequent active/dormant transitions, increasing memory pressure. Evicted states incur costly recomputation upon reactivation, particularly under long-context, short-decode workloads. This…
World models (WMs) generate driving scenarios beyond recorded logs, enabling the training and evaluation of autonomous driving systems in unseen interactions. However, existing WMs struggle to ensure both behavioral realism and visual fidelity, since they overlook the relationship between vehicle interactions and the ego visual observations. We propose a closed-loop interactive WM that couples behavior modeling and visual rendering. First, imitation-learned…
Large language models (LLMs) are often described in terms of personas and this framing has become central to alignment research. In this paper, we argue that there is a deeper level of description that lies in the agent's worldviews, i.e., its beliefs about what reality is and how it works. Treating LLMs as a population of subjects, we administer two worldview surveys from human psychology to eight models. We show that the models share many key worldview positions. For…