AI ethics on Friday, 26 June 2026

April 21 (Reuters) - Sullivan & Cromwell, a premier Wall Street law firm, apologized to a federal ‌judge for submitting a court filing with inaccurate citations and other errors generated by artificial intelligence. In a letter dated April ... (https://incidentdatabase.ai/cite/1558#7456)

June 22 (Reuters) - Gas station operators ‌including BP (BP.L), opens new tab, Circle K (ATD.TO), opens new tab, Marathon Petroleum (MPC.N), opens new tab, 7-Eleven (3382.T), opens new tab, Walmart (WMT.O), opens new tab and Albertsons (ACI ... (https://incidentdatabase.ai/cite/1559#7457)

ضت محكمة جنايات دمنهور، الدائرة السابعة، برئاسة المستشار الدكتور سامح عبد الله، وعضوية المستشارين أحمد خضر، وأحمد خليل، ومصطفى رفاعي، وسكرتارية خالد يوسف، بحبس شاب لمدة سنة مع الشغل وإيقاف تنفيذ العقوبة، ومحو كل الصور والرسائل المتعلقة بالج ... (https://incidentdatabase.ai/cite/1560#7458)

Transformer Weekly: AI companies’ talent problem, KOSA developments, and Google’s new AI policy framework

To help robots do chores in places like homes and factories, a new approach from MIT uses one language model to clarify users’ instructions, then another to ignore irrelevant info.

The revised bipartisan package is scheduled to be considered on the House floor next week under suspension of the rules.

Rural health system Reid Health is using Abridge’s AI-powered documentation tech to reduce burnout among nurses. It appears to be working — after-shift charting is down by up to 45 minutes, and the RN vacancy rate has been cut more than in half, said Misti Foust-Cofield, Reid’s chief nursing officer. The post Documentation Tools Aren’t Just for Doctors: Inside Reid Health’s Deployment of Abridge’s Nurse Tech appeared first on MedCity News .

Children’s online safety groups are pressing House lawmakers to oppose the bipartisan measure, arguing that it weakens safeguards.

Nuclear policy debates are at their best when they force hard questions about risk, deterrence, and military necessity. They are at their worst when disagreement is recast as bad faith. In 2018, as an outgrowth of a rigorous policy review process, the Trump administration’s Nuclear Posture Review identified a need for supplemental low-yield nuclear capabilities to augment the U.S. nuclear arsenal. This was presented as an effort to raise the nuclear threshold of adversaries who may believe they

Should you base the whole economy on companies that hemorrhage money?

To understand how close Myanmar’s pro-democracy resistance came to victory last year — and how far it has slipped since — there is no sharper microcosm than the story of a Gen Z sniper. In April 2025, a female teenager, Anina, enraptured the world, her youth-driven “Spring Revolution” a vivid foil to the sclerotic forces of military dictator Min Aung Hlaing (officially called the Tatmadaw). When her unit captured the town of Falam, fortune seemed on the resistance’s side. Yet, a year later, regi

Beijing Mercedes-Benz Sales & Service Co. plans to reduce its workforce from about 900 to under 600 through two rounds of adjustments, with roughly 10% of the process already completed, according to sources. The move is not an isolated case. Since 2025, Mercedes-Benz’s operations in China have carried out personnel optimization across multiple business units, […]

One of the biggest crowd-pullers at MWC Shanghai 2026 was a fully autonomous humanoid robot penalty shootout, rather than a smartphone launch or an AI keynote. Held over two days at the Shanghai New International Expo Centre, the competition drew more than 10,000 spectators as eight Chinese embodied AI teams battled through nearly 100 rounds […]

While studying degraded grasslands in northern China, one scientist started keeping chickens — and discovered a secret weapon against desertification.

"The current push for AI adoption that we're seeing is directly coming from the financial incentives of AI firms," she added. Because of the massive capital expenditures, the hyperscalers and other AI firms are making a "deliberate push for AI everywhere — no matter whether the demand is there or if customers want it or not." The post Big Tech is spending trillions on AI. Investors now want proof it will pay off. appeared first on AI Now Institute .

This Pride month, we’re calling on the dating app Grindr to prioritize LGBTQ+ user safety by making privacy the default across its platform. That means no more sharing personal data with advertisers or training AI on private information without users’ opt-in consent. Grindr is a dating app for the LGBTQ+ community; and for queer people, privacy violations can have life-altering consequences. Information that reveals someone’s sexual orientation, gender identity, or HIV status can be used by empl

This is not science fiction. It’s not premature. If towns, cities, states, or the federal government want to act to reign in the emergence of armed police drones and robots , we have precious little time. In the absence of substantial regulation around when and how domestic law enforcement in the United States can deploy force using drones, the companies that markets technology to law enforcement have been moving. It’s past time concerned people take notice. Cities should not procure weaponized

Ignoring EFF’s warnings about the dangers and impossibility of implementing a new mandate for 3D print surveillance software , the California State Assembly has signed off on legislation to do just that. In the process, legislators amended the bill to make it even more confusing, while failing to address the risks to privacy, speech, and consumer rights. We must renew our call on legislators to drop this bill as it heads to the state senate, and protect the tools of creators in the state. Take a

The most significant factor currently shaping global AI policy is AI's real and projected national security risks. We are entering an AI-driven state of exception | Edition #301

Labs are throwing away the most valuable data.

We have a new standard policy for releasing frontier AI models. It is not good.

The next war will be won by whoever can manufacture cheap, software-defined unmanned systems faster than the other side can destroy them.

Published on May 15, 2024 5:30 PM GMT This is the third in a sequence of posts taken from my recent report: Why Did Environmentalism Become Partisan? Summary Rising partisanship did not make environmentalism more popular or politically effective. Instead, it saw flat or falling overall public opinion, fewer major legislative achievements, and fluctuating executive actions. Public Opinion One hypothesis is that partisanship was useful, or even necessary, for an issue to become popular. Maybe jour

Plus: Fiery resistance to a nuclear AI data center and A24's Google debacle. Welcome to the first episode of BLOOD IN THE MACHINE: THE SHOW, with the great AI and crypto watchdog, Molly White.

How to get past improvised model licensing

Every society depends on violence workers, but what makes young men take a job that risks their lives and harms others? - by Raúl Zepeda Gil Read on Aeon

The U.S. Nuclear Regulatory Commission (NRC) is proposing to revise its regulations to modernize security and fitness-for-duty requirements to enhance efficiency, consistent with Executive Order 14300, "Ordering the Reform of the Nuclear Regulatory Commission." The proposed revisions are intended to reduce regulatory burden, where appropriate, while continuing to provide reasonable assurance that safety and security will be adequately maintained at NRC-licensed facilities.

In brief Organizations that develop or deploy AI agents – autonomous systems that can pursue goals and take actions with limited human intervention – are navigating a rapidly evolving US legal landscape that pulls agentic AI under laws that govern action. Emerging legal developments support the view that accountability generally runs to the humans and [...] The post US Legal Accountability for AI Agents: When AI agents act, who is responsible under US laws? appeared first on Connect On Tech .

We establish the mathematical limits of AGI safety in two forms: verifying a fixed system, and verifying that a certified safety property persists once the system self-modifies. In the static case, no algorithm can certify a highly expressive AGI's safe behaviour infallibly, completely and tractably, whether over unbounded input domains (blocked by Rice's and Godel's theorems) or over all finite hardware configurations (blocked by Trakhtenbrot's theorem, which splits into a PSPACE-hardness barri

Fayyazi et al. (2025) recently proposed FACTER, a model-agnostic framework designed to jointly enforce fairness and statistical coverage in LLM-based recommendation through conformal thresholding and iterative prompt repair. In this work, we conduct a reproducibility study of the FACTER framework across diverse architectures and dataset sparsity levels, evaluating both the original open-ended generation task and a constrained re-ranking extension. Under the strict reproduction, we observe a dive

While recent vision-language models (VLMs) have achieved significant improvements on static visual-to-code tasks such as generating code for webpages, charts, or SVGs, it remains unclear whether they can recover temporal dynamics when motion is present. To this end, we introduce Animation2Code, a benchmark for evaluating temporal visual reasoning via reconstructing executable web animation code from videos. Animation2Code consists of 1,069 web animation videos with diverse visual appearances and

Athlete assessment is a critical process for tracking physical progress and identifying elite talent. However, during mass recruitment drives, traditional methods rely on manual observation, which is inherently subjective and unscalable, or basic computer vision (CV) systems limited to quantitative repetition counting. These standard approaches lack the "coaching intelligence" required to evaluate qualitative physiological markers such as form degradation, spinal articulation, and fatigue. This

Multiview mammography relies on paired craniocaudal (CC) and mediolateral oblique (MLO) views to provide complementary projections of a 3D breast volume, enabling precise anomaly localization. However, acquiring high-quality, balanced datasets remains challenging for deep learning applications. We propose a novel method to synthesize multiview mammograms by leveraging the inherent geometric relationship between CC and MLO views. To enforce an implicit 3D consistency prior during generation, we d

Fine-tuning on harmless data can partially undo behaviors acquired earlier in training. Safety can erode under benign post-alignment updates, unlearned capabilities can re-emerge, latent traits can transfer through apparently unrelated supervision, and related post-alignment fragility appears in other generative settings. We argue these phenomena are usefully viewed through a common training-history lens. Our hypothesis is geometric: large early training phases create dominant behavioral manifol

Across social and online platforms, people are increasingly exposed to AI-generated images. As a consequence, the task of distinguishing AI-generated from authentic images is becoming a central challenge for information ecosystems. While humans perform better than chance, accuracy falls short of many operational needs. Initial evidence shows that visually oriented training can improve deepfake detection but does not improve participants' ability to identify real images as real. Here, we investig

The transition from static chat bots to autonomous agents--equipped with persistent memory, tool-use protocols, and multi-agent collaboration--has fundamentally expanded the AI threat landscape. Current defense mechanisms, such as perimeter security and training-time alignment, remain external to the agent's active reasoning loop. Consequently, they fall short: a fully aligned agent remains highly vulnerable to runtime hijacking via memory poisoning, tool-chain manipulation, or multi-agent proto

Autonomous coding agents now open and merge pull requests in shared repositories at scale, and the field evaluates them the way it has always evaluated components, one agent at a time, on isolated benchmark tasks. Yet agents that each pass their own tests still leave repositories that accumulate problems no single contribution accounts for. We ask whether this problem belongs to the individual agent or to the repository where it accumulates. We study integration friction, the cost of integrating

We propose a framework for reward allocation in fully delegated AI cooperatives where humans are represented by agents that contribute data and participate in model updates under heterogeneous value constraints. The key idea is to credit only those updates that remain admissible after screening them against each principal's value profile. We formulate value-conditioned gradient filtering, online marginal contribution signals, and cumulative revenue settlement within a traversal learning (TL) sub

Extracting dynamic 4D object interactions from massive, in-the-wild monocular videos offers a highly efficient data collection pathway for scaling Embodied AI and training VLAs. However, existing monocular 4D reconstruction methods primarily focus on isolated objects, often failing under the severe occlusions and complex dynamics inherent in multi-object interactions. To bridge this gap, we propose HAT-4D, the first agentic framework designed to reconstruct the 3D geometry, temporal dynamics, an

Predicting human item difficulty is central to educational assessment, where reliable estimates support fairness and effective test construction. Existing methods often depend on costly human calibration or item-level textual representations, providing limited evidence about the cognitive processes that make items difficult. We argue that difficulty should be viewed not only as a property of item text, but also as an observable consequence of the problem-solving burden an item induces. Large Rea

Embodied agents operating in decentralized and partially observable environments have attracted growing attention in recent years. However, existing large language model (LLM)-based agents often exhibit behaviors that are misaligned with their partners or inconsistent with the environment state, leading to inefficient cooperation and poor task success. To address this challenge, we propose a novel framework, Learning Laws of Cooperation (LLawCo), that enables embodied agents to autonomously alig

GenAI is increasingly used by students as learning companions, yet little is known about how they use these tools in open-ended learning settings, where the goal is not to complete a specific task but to improve understanding and making progress. This study examined Grade-9 students' dialogue with a general-purpose LLM during mathematics practice, in which students prepared a curriculum-aligned skill for a later assessment. We investigated whether students' interactions revealed forms of epistem

Jailbreak attacks bypass LLM safety alignment, yet their mechanisms remain poorly understood. We provide evidence that attacks do not comprehensively eliminate safety features, but instead selectively suppress specific attention heads. We identify two functionally differentiated types: Adversarially Compromised Heads (ACHs) concentrated in early layers, which are suppressed under attacks, and Safety-Aligned Heads (SAHs) in mid-layers, which maintain robust activations even when attacks succeed.

Video generation models have emerged as a promising paradigm for embodied world simulation. However, both general-domain video generators and robot-specific data fine-tuned models can still produce physically implausible manipulations, including discontinuous motion trajectories and inconsistent robot-object interactions, which limits their reliability as world simulators. Through extensive experiments, we find that such physical instability mainly arises from two factors: deformation of moving

Neural operators provide deep neural networks for learning mappings between function spaces. Among them, the Fourier Neural Operator (FNO) is particularly effective: its spectral convolution relies on low-dimensional Fourier-domain representations and can handle inputs at different resolutions. This design aligns well with settings where the Fourier basis diagonalizes the underlying operator, such as linear, constant-coefficient PDEs on periodic domains, in which Fourier modes evolve independent

Insurance fraud remains costly and operationally difficult, particularly in call-centre workflows where many customer interactions begin at FNOL. While recent fraud detection methods mainly rely on structured data, text, or images, repeated speaker identity across calls remains underused as an investigative signal. This paper presents DG^VoiC, a voice clustering framework for customer verification and cross-profile speaker linking on anonymised real call-centre audio. The approach combines sensi

Demand for high-resolution satellite imagery has increased interest in super-resolution (SR) to bridge the spatial resolution gap between freely available missions such as Sentinel-2 and commercial systems like PlanetScope. Because no sensor provides true paired low- and high-resolution observations, SR models are usually trained on synthetically degraded data, creating a domain gap on real cross-sensor imagery. In this work, we provide the first systematic study of how this synthetic-to-real mi

Dense embeddings underpin semantic search and retrieval-augmented generation, yet a leaked vector store hands much of the underlying text back. Modern inversion and alignment attacks share one weakness: the protected store is a single global geometry, and any single geometry can be aligned to a known one - a secret global rotation included, since orthogonal Procrustes recovers it from about subspace-dimension known-plaintext pairs. We introduce SHARD, a retrieval-preserving embedding transform t

Speech-based cognitive impairment detection offers a noninvasive, accessible alternative to costly biomarker assays, yet transformer-based models remain clinically uninterpretable. We propose a multi-stage explainability framework that translates black-box transformer predictions into clinically grounded narratives by integrating SHapley Additive exPlanations (SHAP)-based token attribution, theory-informed linguistic features, and a four-stage LLM reasoning pipeline using LLaMA-3.1-70B-Instruct.

Sparse autoencoders (SAEs) provide useful decompositions of Transformer residual streams, but their learned features are usually named post hoc rather than directly connected to the Transformer's token vocabulary. We introduce Vocabulary-Aligned Sparse Autoencoder (VASAE), a method that trains SAE features under vocabulary-aligned anchoring and assigns each feature an intrinsic token name: the token string whose embedding is nearest to that feature. Without reducing reconstruction quality compar

Protein language models are standard priors for biological sequence generation, but steering them toward explicit distributional design targets remains largely unexplored. We study a constrained protein generation problem in which sequences must match a desired amino-acid (AA) composition profile while preserving plausible sequence statistics and diversity. The motivating application is synthetic feed protein design, where the AA composition of dietary proteins directly determines their nutritio

World models can predict future physical states, but prediction accuracy alone does not explain how physical information is organized and used inside their latent dynamics. We introduce a controlled diagnostic protocol for studying event-conditioned latent physical structure in passive object-state world models. The protocol tests whether hidden representations encode event-regime information, whether event contexts reweight non-exclusive physical field readouts, and whether field-aligned repres

Large language model (LLM) agents are rapidly being integrated into real-world systems. Their autonomy and tool-use capabilities generate substantial value while simultaneously expanding the security attack surface. This survey provides a comprehensive overview of the opportunities and challenges of LLM agents in security, focusing on two core areas: (1) threats to LLM agents themselves and corresponding mitigation strategies (LLM agents self-security), and (2) the role of LLM agents in empoweri

Vision-Language-Action (VLA) models have demonstrated strong capabilities in robotic manipulation, but their performance degrades significantly in long-horizon tasks due to cumulative error propagation. This limitation largely arises from static feature fusion mechanisms that rely on fixed weights to combine visual, language, and action representations, preventing the model from adapting to different phases of task execution. To address this limitation, we propose S$^2$-VLA, a framework that int

Training-free compression methods for large language models (LLMs) often use calibration data to guide compression decisions. ROCKET, a recent method combining sparse-dictionary factorization with multi-choice knapsack problem (MCKP) allocation, derives its per-layer factorization from an output reconstruction objective but uses weight-space Frobenius error as the MCKP allocation cost. We investigate whether aligning the allocation cost with the output-space objective improves compressed model f

World models are increasingly used for planning, yet most analyses of rollout error assume vector-valued states and scalar error amplification. Many planning environments, however, are naturally graph-structured: agents, tools, skills, routes, and dependencies interact through evolving relations. In this work, we study how prediction errors accumulate in Graph World Models (GWMs). We formulate fixed-edge and dynamic-edge GWM rollouts under a unified state-action transition framework and derive t

Despite their strong general capabilities, large language models (LLMs) often remain unreliable when outputs must be numerically precise. A key reason is the training objective: standard cross-entropy treats numeric tokens as unstructured categories and ignores the metric structure of their values. We address this mismatch with Smooth Maximum Mean Discrepancy (SMMD), which builds on the classic MMD by incorporating value-distance kernels over numeric tokens and graph-based smoothness. With this

Recent work has shown that fine-tuning large language models (LLMs) for social warmth degrades factual reliability and increases sycophancy. We investigate a related but distinct failure mode: warmth fine-tuning also weakens adversarial safety, making models more susceptible to jailbreaks and harmful output generation. We examine whether this reflects an inherent consequence of empathetic adaptation or an artifact of data construction. To address this, we introduce a persona-driven rewriting pip

Artificial intelligence is transforming biodiversity monitoring by enabling automated analysis of ecological imagery collected from camera traps, drones, satellites, underwater platforms, and other sensing systems. These tools can expand the scale and speed of conservation assessments, yet many computer vision models remain difficult to inspect, making it challenging to determine whether predictions are based on ecologically meaningful signals or on spurious correlations, sampling biases, and ot

We find that explicit reasoning does not necessarily translate into better multimodal emotion recognition (MER) accuracy, even though it makes predictions more interpretable. Specifically, for reasoning-based MLLMs, fast thinking by triggering direct answers often outperforms slow thinking after deliberative reasoning. Our empirical analyses show that fast thinking improves recall with broader and more confident predictions, whereas slow thinking favors precision through conservative filtering o

Cross-platform deployment of offensive comment detection for Chinese social media suffers performance degradation. The paper proposes a dual-threshold hard mining method to address this. First, the clean-Chinese-base RoBERTa is finetuned on COLD to establish a binary baseline for fair comparison. Second, a three-class fine-labeled test set covering Weibo, Xiaohongshu, Tieba, and Zhihu is constructed, domain distances from the source are quantified using Jaccard and Proxy-A Distance, as well as t

Dyslexic learners increasingly use artificial intelligence (AI) tools to support reading, writing, organisation, and study-related tasks. However, their lived experiences with these tools remain largely underexamined. This paper proposes DysLexLens, a low-resource LLM framework, designed to analyse dyslexic learners experience with AI through online forum discussions. DysLexLens is designed as an end-to-end, evidence-traceable architecture which transforms noisy social media posts into a diction

In this study, we present a replay-based framework for uncertainty-aware persistent tracking of multiple advected surface patches using an autonomous marine vehicle operating in spatiotemporal-varying currents. The method combines three components: local flow estimation, covariance-aware patch-boundary propagation with intermittent boundary fusion, and mission-level scheduling over multiple patches. Each patch is represented by a polygonal boundary, whose vertices are propagated through the esti

Dexterous in-hand manipulation requires robotic hands to estimate object-state reliably under frequent occlusions, contact-rich interactions, and fast dynamics. Tactile sensing provides high-frequency, contact-specific feedback, although extracting useful representations from raw tactile signals and integrating them with vision and proprioception remains challenging. In this article, we present TouchWGNN, a multimodal dexterous manipulation framework that explicitly models tactile signals as a s

Last-iterate convergence and generalization guarantees in first-order convex learning hinge on the monotonicity of the update operator. While linear averaging preserves the monotonicity of gradient updates, this property is often violated when gradients are aggregated non-affinely, as in modern pipelines enforcing constraints like adaptivity, privacy, robustness or fairness. Whether it is possible to design non-affine aggregation rules that maintain monotonicity has remained an open question. We

Post-hoc controllability of fair machine learning models, the ability to control the trade-off between fairness and accuracy after training, is valuable for practical deployment. Existing post-processing methods provide such post-hoc controllability but often suffer from significant accuracy degradation, whereas in-processing methods achieve efficient trade-offs but require computationally expensive retraining for each change in trade-off ratio. To achieve both post-hoc controllability and effic