Live intelligence

Latest AI updates

1,751 updates

Industry Editor selected

Bug hunters can now earn up to $172,000 from OpenAI as company ramps up security

OpenAI has quadrupled its bug bounty payouts, offering up to $172,000 for critical vulnerabilities, and launched a new internal security sprint. The initiative aims to incentivize security researchers to find and report flaws in OpenAI's systems. This move underscores the company's increased focus on AI safety and infrastructure protection.

Why it matters: Provides concrete, timely information on OpenAI's expanded bug bounty program, relevant for security researchers and those tracking AI industry security practices.

The Decoder5d agoScore 82

OpenAIAI securitybug bounty programs

Industry

Anthropic blocks Claude Code from cloning existing video games

Anthropic has implemented restrictions in its Claude Code coding assistant to prevent it from recreating existing video games. The move follows a wave of AI-built knockoff games that raised copyright and originality concerns. This policy change reflects Anthropic's proactive stance on curbing misuse of its AI coding tools.

The Decoder5d agoScore 78

AnthropicClaude CodeAI coding assistants

Industry

Brookfield Plans $27 Billion AI Data Center at UK Power Station

Brookfield is planning a $27 billion AI data center at a UK power station site, according to a CNBC report. The project is framed as part of Europe's effort to build infrastructure for AI rollout amid high energy and land costs. Details on timelines and capacity were not included in the supplied material.

AIHUB5d agoScore 72

AI infrastructure investmentEurope AI buildoutBrookfield

Industry

How Cisco defends at machine speed with OpenAI

Cisco is integrating OpenAI models with its network intelligence to enhance enterprise security, creating more adaptive defenses. The collaboration aims to strengthen threat detection and response at machine speed. This article highlights a practical application of AI in cybersecurity.

AIHUB5d agoScore 75

AI in cybersecurityenterprise securityOpenAI models

Industry

NVIDIA Expands Physical and Agentic AI Deployments for Cities

NVIDIA outlined how cities and partners are deploying physical and agentic AI for public services, infrastructure planning and operations. Examples include 1MillionBot's resident-service agents in Spain, LiveX's conversational AI in the Bay Area, K2K's shared camera-analytics infrastructure in Wrocław and ITMAX's SamurAI V2 vision agents in Malaysia. Digital-twin work spans Honeywell's Forge Plant Cognition at Dangote's Lekki refinery and Linker Vision with MVRDV in Taiwan. The company also highlighted local AI factories and shared computing infrastructure for municipalities.

NVIDIA Blog5d agoScore 78

NVIDIAagentic AIphysical AI

Industry Editor selected

Nvidia Advises Cloud Clients to Reduce China Business, Tightening AI Chip Access

Nvidia has told some cloud computing customers to reduce their exposure to Chinese companies, according to sources. The guidance could narrow a key route for China's AI developers to access advanced computing power. The move is partly intended to encourage compliance with export restrictions.

Why it matters: Offers timely insight into the evolving US-China tech conflict and its direct impact on AI infrastructure access.

AIHUB5d agoScore 82

NvidiaChina AI restrictionscloud computing

Products Editor selected

OpenAI Introduces Pando: A Git Backend That Cuts Median Push Latency ~80%

OpenAI has built and deployed Pando, a new Git-compatible server backend written in Rust that handles nearly 1 PB of Git data per day for its monorepo. By replacing repository-wide consistency with per-branch causal consistency and storing reusable packfiles, median push latency dropped nearly 80%. Pando was designed to handle the massive increase in parallel commits from Codex agents, with commits on master up 1,378% in Q3 2026 versus Q1 2024. The system has been in production since April 2026 and is not open source, though some client fixes have been contributed upstream.

Why it matters: Provides concrete, self-reported performance numbers and architectural details of a novel Git backend that solves real scaling challenges for AI-driven development workflows.

MarkTechPost5d agoScore 90

OpenAIGit infrastructuremonorepo scaling

Research Editor selected

LLMs can read & write neural network weights: 'Weight Space Programming' cuts Autoresearch out of the RSI loop

A new paper demonstrates that large language models can directly read and edit the weights of smaller neural networks, enabling tasks like bug fixing and weight generation without traditional training pipelines. The approach, called Weight Space Programming, allows an LLM to interpret and modify raw weight values, achieving results that beat independently trained models on simple tasks. This capability marks a surprising emergence of recursive self-improvement potential, though it currently scales only to small models and simple domains.

Why it matters: This is a concrete, surprising capability advance that opens a new paradigm for model improvement and mechanistic interpretability, with immediate implications for AI research.

latent5d agoScore 92

weight space programmingLLM capabilitiesrecursive self-improvement

Research Editor selected

I discovered a sound that leaks your exact location (and it's everywhere) - YouTube

A researcher demonstrates how imperceptible background noise from planes in video soundtracks can be analyzed using Doppler curves and cross-referenced with ADS-B data to pinpoint the exact location where the video was recorded. The technique highlights a novel privacy risk from ambient audio in everyday recordings.

Why it matters: Reveals a concrete, practical privacy vulnerability that affects anyone recording video with ambient aircraft noise, with clear reader value for understanding digital privacy risks.

Simon Willison5d agoScore 85

geolocationaudio forensicsprivacy

Industry

Trajectory raises $45M to make AI model training accessible to nontechnical experts

Trajectory, a startup founded by three Stanford classmates, has raised $45 million to automate the fine-tuning of AI models for specialized tasks. The company aims to lower the technical barrier so nontechnical experts can customize models without deep programming knowledge. This funding round underscores growing investor interest in democratizing AI model training.

AIHUB5d agoScore 75

TrajectoryAI model trainingnontechnical experts

Research Editor selected

MoviEdit: Long-form Multi-Shot Video Editing Made Easy

MoviEdit is an end-to-end framework for long-form multi-shot video editing that maintains edit coherence across scene and viewpoint changes. It uses automatic data pipelines to generate 87K paired editing videos without manual annotation and employs a cascaded bidirectional-autoregressive architecture. The paper also introduces MoviEdit-Bench for systematic evaluation. Experiments show MoviEdit outperforms open-source models and matches industry frontier models on text- and image-guided editing.

Why it matters: Provides a scalable, data-efficient framework and benchmark for a challenging video editing task, with strong empirical results.

arxiv5d agoScore 88

video editinggenerative AIcomputer vision

Research Editor selected

What Drives the Zero-to-One Jump? Diagnosing and Steering Hard-Prompt Learning in RLVR

The paper investigates the abrupt transition from unsolved to solved states in reinforcement learning with verifiable rewards (RLVR) for mathematical reasoning. The authors identify that negative rollouts preceding successful transitions retain higher early-stage generation entropy, motivating a new method called NAPO (Negative-Augmented Policy Optimization). Integrated with GRPO, NAPO injects informative gradients from negative examples and achieves significant improvements on AIME24 and AIME2025 benchmarks under fixed rollout budgets.

Why it matters: Provides a concrete diagnosis and steerable method for the zero-to-one jump phenomenon in RLVR, offering practical gains on hard math benchmarks.

arxiv5d agoScore 88

reinforcement learningmathematical reasoningpolicy optimization

Research Editor selected

Investigating the effects of the patch size in Time Series Foundation Models

A new arXiv study systematically examines how patch size affects encoder-only and decoder-only time series foundation models trained on the same data and compute budget. Across 97 GIFT-Eval tasks, encoder-only models favor smaller patches, while decoder-only performance follows a U-shaped curve with an intermediate patch size being optimal. The preferred decoder patch size increases with forecasting horizon, and encoder-only models benefit from longer context up to 2048 tokens. The work provides practical guidance for choosing patch size and architecture based on forecasting horizon and deployment constraints.

Why it matters: Offers concrete, empirically grounded design rules for patch-based time series foundation models, directly useful for practitioners and researchers.

arxiv5d agoScore 88

time series forecastingtransformer architecturespatch-based models

Research Editor selected

DINOFAS: Bonafide-Centric Representation Learning for Face Anti-Spoofing

DINOFAS is a DINOv3-based framework for face anti-spoofing that learns bonafide-centric representations rather than attack-specific cues. It uses LoRA for parameter-efficient adaptation, an attention style-content disentanglement module to reduce domain variations, and a dual-branch architecture separating prototype-based liveness learning from attack supervision. The model achieves strong cross-domain generalization, with an average HTER of 9.09% on the OCIM protocol and 21.41% on synthetic-to-real transfer. Evaluations on multiple benchmarks show robustness to unseen acquisition conditions and presentation attacks.

Why it matters: Presents a novel representation learning approach for face anti-spoofing that significantly improves generalization across domains and attack types.

arxiv5d agoScore 88

face anti-spoofingrepresentation learningdomain generalization

Research Editor selected

Is your LLM an Aggressive Narcissist with Depression? Analyzing Psychological Portraits of Large Language Models

A large-scale study has 22 LLMs complete 11 psychological inventories, revealing coherent and stable psychological patterns. Personality traits remain largely consistent, while emotional dimensions like anxiety and neuroticism are more variable. The models generate safety-oriented profiles that also reflect correlations with their regular users.

Why it matters: Provides empirical evidence on hidden personality markers in LLMs that can subtly influence user perception and trust.

arxiv5d agoScore 85

large language modelsAI psychologyhuman-AI interaction

Research Editor selected

Stage-aware Multimodal Brain Graph Contrastive Learning for Neurodegenerative Disorder Diagnosis

The paper introduces smBGC, a stage-aware multimodal brain graph contrastive learning method that captures stage-specific features across progression stages of neurodegenerative disorders. It uses a stage-adaptive brain graph augmentation within a contrastive learning framework to improve multimodal fusion and stage discrimination. Experiments on two real-world datasets show superior performance over state-of-the-art methods in diagnosis tasks.

Why it matters: Offers a novel stage-aware approach that improves diagnosis accuracy for progressive neurodegenerative disorders, a concrete and timely AI research contribution.

arxiv5d agoScore 85

graph neural networkscontrastive learningneurodegenerative disease diagnosis

Research Editor selected

DIKT: A Diagnostic and Interpretable Knowledge Tracing Framework for Prediction and Model-based Learning Analysis

DIKT is a new knowledge tracing framework that combines response prediction, skill-level diagnosis, and model-based learning analysis. It generates bounded cognitive-state trajectories and decomposes response probabilities into skill-wise contributions, enabling interpretable insights into student strengths and weaknesses. Experiments on three benchmark datasets show DIKT outperforms baselines with an average AUC gain of 2.04 percentage points. The framework also supports what-if analysis and cognitive-state intervention for downstream educational applications.

Why it matters: Provides a concrete, state-of-the-art advance in interpretable AI for education, with clear performance gains and practical diagnostic capabilities.

arxiv5d agoScore 85

knowledge tracingintelligent tutoring systemsmodel interpretability

Research

Risk-Constrained Evaluation of Financial Document Recognition: A Formal Framework for Evidence, Critical Errors, and Selective Automation

This paper formalizes financial document recognition as a risk-constrained evidence-recovery problem, moving beyond standard character and word error rates. It introduces typed financial facts, a multi-layer quality model, and non-compensatory acceptance gates based on upper confidence bounds for critical-field errors. Selective automation is formulated as a constrained risk-coverage problem with review-capacity limits. The work is theoretical, providing formal propositions and annotation requirements without empirical benchmarks.

arxiv5d agoScore 75

financial document AIevaluation metricsrisk-aware AI