SnapLogic launched SnapCode and a Model Context Protocol (MCP) server that enables AI coding tools like Claude Code to directly access over 1,000 pre-built enterprise integrations, including SAP, Salesforce, and Snowflak…
A software auditor reveals that enterprise AI contracts using per-seat pricing are a scam, with 90% of employees rarely using the tools. The model subsidizes vendor profits on unused compute, akin to gym memberships. The…
OpenAI launched a new ChatGPT search feature on July 14, 2025, allowing users to search across all their conversations, projects, images, and documents, available on web, iOS, and Android for all plan tiers. The update f…
Researchers introduced Weight-Adjusted Gradients (WAG), a novel method for identifying critical parameters in large language models that disproportionately influence behavior and performance. WAG reveals a tiny subset of…
Researchers have developed RecursiveMAS, a framework for multi-agent AI collaboration that uses recursive loops to improve accuracy by 8.3%, boost inference speed by up to 2.4 times, and reduce token usage by 34.6% to 75…
New York Governor Kathy Hochul announced a one-year moratorium on state environmental permits for new hyperscale data centers, making New York the first state to take such action. The pause applies to facilities using 50…
A new study reveals that existing class unlearning methods in machine learning overlook class geometry, enabling Class Membership Inference Attacks (CMIA) that leak information about forgotten data. Researchers propose T…
A new study on robust AI planning introduces non-asymptotic convergence bounds for Q-learning and actor-critic algorithms in distributionally strong Markov Decision Processes, achieving an optimal strong policy with accu…
The Bank of England plans to ease capital rules to spur lending for AI investment, even as Governor Andrew Bailey warns of a 'triple whammy' of risks from frothy AI stocks, slow adoption, and rapid development. Meanwhile…
Elon Musk's AI chatbot Grok, integrated into X's automatic translation feature, is generating profane and sexually explicit mistranslations of innocent user posts, such as describing a coffee-making video as public mastu…
Anthropic released a new commercial that juxtaposes tough questions about AI trust, job loss, and societal impact with optimistic visions of the technology's potential, created by ad agency Mother London. The ad reflects…
A developer built Gubernator, a lightweight container orchestrator in Go that combines Docker Compose simplicity with Nomad-like decentralized targeting. The project evolved from a single-node API to a multi-node VM clus…
Twenty-six former Meta employees filed a federal lawsuit in Oakland, California, alleging the company's AI-powered performance tools systematically penalized workers on medical or parental leave during the 2026 layoff cy…
Spet, a free no-signup transcription service, launches with strong accuracy on non-English audio using OpenAI's Whisper large-v3 model. The platform supports 90+ languages, deletes audio after transcription, and handles …
Twenty-six current and former Meta employees filed a federal lawsuit Monday accusing the company of using AI-powered systems, including keystroke monitoring and a large language model assistant called Metamate, to dispro…
Bonsai AI released Bonsai 27B, the first 27-billion-parameter model capable of running on a phone, using 1-bit and ternary weight compression to achieve a 3.9 GB footprint. The model retains 90-95% of full-precision perf…
Anthropic launched India-specific pricing for its Claude AI assistant on July 14, 2026, targeting the country's price-sensitive developer market of over 5 million professionals. The move aims to compete with OpenAI and G…
Meta's SAM 3 model enables zero-shot text-prompt segmentation, allowing users to segment objects like "helmet" or "safety vest" without training or clicks. A Roboflow Workflow tutorial demonstrates using SAM 3 for constr…
Google DeepMind CEO Demis Hassabis called for the creation of an independent standards body to regulate frontier AI models, modeled after the Financial Industry Regulatory Authority (FINRA). The proposed regulator would …
Researchers found that training language models to be risk-averse on low-stakes gambles (up to $100) can generalize to astronomically high stakes (up to $10^98), with Qwen3-8B choosing a safe 'Cooperate' option rising fr…