Signal Over Noise.

The White House has ordered Anthropic to halt exports of its advanced AI models, Fable and Mythos, raising questions about the effectiveness of using decades-old export control tactics on cutting-edge software.

The US government has ordered AI company Anthropic to pull its new Fable 5 and Mythos 5 models, sparking a debate over national security, industry-wide vulnerabilities, and the ban's potential to inadvertently boost the company's brand.

OpenAI introduces new enterprise-grade usage analytics and spend management tools, giving organizations greater control and visibility over AI platform costs.

OpenAI's platform is experiencing service disruptions with repeating verification prompts, signaling a potential backend overhaul aimed at improving ChatGPT's health intelligence capabilities.

Google has released its AI Control Roadmap, a new security framework that treats advanced AI agents as potential insider threats and adds system-level controls on top of traditional model alignment.

ServiceNow researchers have released MosaicLeaks, a new benchmark and training method that drastically reduces privacy leaks from AI research agents by training them to construct safer web queries without sacrificing performance.

New benchmarks from Hugging Face reveal that while the popular LoRA fine-tuning method is effective, alternative PEFT techniques like OFT and BEFT can offer superior performance and memory efficiency, challenging LoRA's default status in the AI community.

The U.S. Commerce Department alleges a critical ASML EUV chipmaking machine may have illicitly reached China, a claim the Dutch technology giant firmly denies, escalating tensions in the global semiconductor supply chain.

Enterprise software company Elastic has reportedly agreed to acquire AI SRE startup DeductiveAI for up to $85 million to bolster its observability platform with automated bug resolution technology.

An AI-powered robotic system demonstrates near-autonomous capabilities by successfully optimizing a difficult chemical reaction crucial for medicinal chemistry and drug discovery.

A new benchmark, LifeSciBench, has been introduced to rigorously evaluate the performance of large language models on complex tasks within the life sciences domain.

The Allen Institute for AI (AI2) has released MolmoMotion, a new model that forecasts 3D object motion from text commands, along with a large-scale dataset and evaluation benchmark to advance research in robotics and controllable video generation.

Google's recent integration of Gemini AI into Google Docs is prompting users to seek methods for disabling the features, highlighting a growing tension between AI assistance and user workflow control.

Former Sequoia Capital leader Roelof Botha joins the board of the newly public SpaceX, bringing extensive financial governance experience to a company where CEO Elon Musk maintains near-absolute control.

Major AI labs are increasingly using advanced simulation environments to test and predict the real-world behavior of AI models and agents before their public release.

AWS's Strands Robots SDK now integrates LeRobot, offering a unified, open-source workflow to move from simulated data collection to policy deployment on physical hardware using the Hugging Face Hub.