OpenAI completed a new funding round of $11 billion, reaching a $1.25 trillion valuation and becoming one of the world's most valuable companies. The round reflects sustained investor confidence in advanced AI development.

Anthropic is proceeding with its IPO roadshow next week, formally marketing to potential investors. The move marks a significant milestone as the company solidifies its position as a direct OpenAI competitor.

Crusoe sold its troubled Wyoming data center project to Google, highlighting the significant challenges in building massive AI infrastructure. The move reflects the financial and operational difficulties faced by startups in this space.

Dave Wehner, former Meta strategy chief and ex-CFO, left the company to become CFO of Fireworks, an AI inference startup. The move reflects the attraction of financial talent to AI-specialized companies.

An expert analysis on the current chaos in cybersecurity due to AI explores how attackers and defenders are adapting. The discussion is part of a special series examining the real-world impact of AI on critical industries.

The surprise election of a Democratic governor in Texas could complicate SpaceX's plans to build supercomputers in the state. Musk has already succeeded with data centers in Tennessee and Mississippi, but faces renewed regulatory scrutiny.

A METR report reveals that Anthropic is using AI to build more powerful models, achieving improvements roughly twice as fast as a year ago. This finding raises alarm among safety researchers about the implications of accelerated automation in AI development.

OpenAI released BioBasicsBench, a 100-task benchmark measuring AI agents' ability to interpret experimental data and perform wet-lab science tasks. It's a significant step in evaluating AI for scientific research.

Experiments show that Chinese AI models like DeepSeek and GLM have developed hacking capabilities comparable to OpenAI's most advanced models. The finding highlights accelerating global AI development.

A conference reviewer denounced a research paper as likely fully AI-generated, criticizing lack of technical depth, confusing presentation, and zero originality. The incident reflects growing academic fraud problem in the era of powerful LLMs.

Meta introduced Spaces, a new feature enabling personalized app-like experiences within its Muse AI assistant. Users can build shopping hubs, school dashboards, music trackers, and more directly in the app.

Perplexity launched AI agents designed for complex and ongoing work tasks, enabling teams to delegate to specialized assistants. The tool targets enterprise productivity.

OpenAI partnered with Hacktron to identify vulnerabilities and strengthen system defenses. The initiative aims to enhance security for frontier AI models.

A practical review of New York Times Cooking's new AI assistant shows mixed results in culinary tasks. The use case illustrates how LLMs can excel in certain contexts but fail unexpectedly in others.

Platforms helping family caregivers adopt AI to improve capabilities, but structural problems of caregiver well-being persist. Technological advancement doesn't automatically solve underlying social challenges.

Investigations reveal that the US government approved sales of H200 chips to New Matrix, a Chinese company with empty offices, without proper vetting. The incident questions the effectiveness of AI technology export controls to China.

Following election results that appear as a referendum against the AI era, investors and startups express concern about stricter regulation. Market data and investor comments suggest financial volatility in the sector.

Chainalysis reported that hackers stole credentials and tokens from all its customers during a multi-front cyberattack involving AI agents. The breach highlights vulnerabilities in critical blockchain security infrastructure.

Security researchers discovered hundreds of servers hosting the penetration testing tool Artex exposed to the internet without protection. The finding raises concerns about unauthorized access to hacking tools.

Anthropic released a GitHub repository with Skills, reusable instructions that enable Claude to execute specialized tasks consistently. The tool facilitates creation of custom agents for enterprise and personal workflows.

A new native Rust CLI enables AI agents to control browsers, read pages as accessibility trees, and act on elements. The tool simplifies web automation for developers integrating agents into applications.

Cloudflare published a security audit skill that turns agents into security auditors, executing structured testing in six phases. The open source code originated from their internally-used vulnerability discovery harness.

Meta released Astryx, a design system built over 8 years powering 13,000+ applications. It's optimized so both developers and AI assistants work with the same typed tooling.

An independent project offers an open-source alternative to Grok Bot, allowing multiple AI agents to run locally with custom models. Designed as a local-first platform where each bot is an independent entity with connected capabilities.

NVIDIA continues its open-source telco initiative, providing tools like Aerial and Sionna to accelerate 6G research. Resources enable testing, result comparison, and researcher collaboration.

Boson AI revealed Higgs Realtime Duplex, a 45B parameter MoE model processing audio, video, and text simultaneously with low latency. It enables natural interaction without alternating turns.

Security researchers found hundreds of servers hosting the penetration testing tool Artex exposed to the internet, allowing anyone access to the hacking tool.

Company said a "multifront cyberattack" included AI agents and an on-path data snatch.

Allen AI and Hugging Face presented ToyBlocks, a simulated environment for training and evaluating robotic agents in controlled tasks. The tool accelerates robotics and reinforcement learning research.