AI Research

The US government reversed export restrictions on Anthropic's Fable 5 and Mythos 5 AI models, allowing global access after nearly three weeks. The restrictions were triggered by cybersecurity concerns, but after collabor…

Oracle spent $55.7 billion building AI data centers this fiscal year, with plans to spend even more by 2027, but the widening gap between capital expenditures and cash flow has become Wall Street's clearest argument that…

A rare bipartisan AI bill, the CREATE AI Act, is advancing through Congress to permanently establish the National Artificial Intelligence Research Resource (NAIRR) at the National Science Foundation. The bill aims to exp…

Brian Patrick, founder of Panodime, runs a daily social media series called AI Executive Insanity Series that highlights outlandish statements from AI executives, aiming to educate the public and advocate for democratic …

AI strategist Sol Rashidi, chief strategy officer at Cyera and a senior fellow at Harvard Kennedy School, told Business Insider she deactivated half of her autonomous agents because they were unreliable, spending more ti…

Andrew Ng introduces three key loops for building 0-to-1 products with AI agents: the agentic coding loop, developer feedback loop, and external feedback loop. He argues that humans retain a context advantage over AI, ma…

Anthropic published an article on recursive self-improvement, exploring how AI systems may increasingly contribute to building better versions of themselves. A developer argues that as AI writes, tests, and reviews code,…

Researchers at a recent study argue that AI systems in the public sector are often oversimplified into a single category, leading to imprecise evaluations of their impact on public values. They propose a five-category ty…

Large language models are evolving from text generators into integrated components in diverse systems, presenting new security challenges across their entire lifecycle. Experts warn that vulnerabilities span eight critic…

ViToS, a dual-stream reinforcement learning framework, prunes unnecessary visual tokens in medical imaging to streamline vision-language models. It reduces inference load and boosts performance, achieving up to 108.27% r…

Anthropic's AI system Mythos, previously banned by the US government over national security risks, is returning with restrictions. The US lifted the export ban after Anthropic worked with the Commerce Department to enhan…

A new study finds Random Forest achieves a 0.99 F1-score in detecting intrusions on IoT devices using the Gotham2025 dataset, outperforming XGBoost, Logistic Regression, Naive Bayes, and Deep Neural Networks. The researc…

A new study compares two deep learning frameworks for uncertainty estimation in automotive radar: a von Mises ensemble (ENS) and an evidential deep learning (EDL) model. The ENS model offers better precision under normal…

Researchers introduced MPL-MAE, a new 3D autoencoding framework that reduces positional bias to enhance semantic learning in point cloud processing. The framework uses a gated positional interface module to balance spati…

Researchers have developed a CVE-TTP Knowledge Graph linking software vulnerabilities to attacker behaviors using the MITRE ATT&CK framework. Transformer-based models like CySecBERT achieved high F1-scores for identifyin…

Researchers introduced Stage-Transition Dense Reward (STDR), a framework that extracts dense rewards from unstructured expert videos to train robotic manipulation agents, improving sample efficiency and success rates in …

AI strategist Sol Rashidi fired half of her AI agents after spending more time supervising them than doing useful work, a phenomenon called 'botsitting' that a Glean report says costs white-collar workers an average of 6…

A researcher used Anthropic's Claude Opus 4.7 to exploit a vulnerability in Front Gate, the ticketing platform for major festivals like Lollapalooza and Bonnaroo, enabling the issuance of free tickets. The breach raises …

OpenAI's GPT-5.6 Pro tier may launch in three distinct variants, according to a benchmark paper, marking a shift from the company's single top-tier model strategy. The move aims to cater to specialized needs and stay com…

GPT-3 is not AGI but language modeling differs from other AI domains because natural language encodes a world model. Scaling language models may drive loss toward Shannon entropy, yielding a usable world model that can b…