AI News Roundup: Anthropic’s $100M Enterprise Bet, GTC 2026 Preview, Washington Passes Chatbot Safety Bill

Anthropic Pours $100M Into Enterprise Push While Fighting the Pentagon

Anthropic is playing offense and defense simultaneously. On Wednesday, the company launched the Claude Partner Network with $100 million in 2026 funding, enlisting Accenture, Deloitte, Cognizant, and Infosys to make Claude the default AI platform for global enterprises. Accenture alone is training 30,000 professionals on Claude, while Cognizant has opened access across its entire 350,000-person workforce.

The enterprise push comes as Anthropic wages an unprecedented legal battle against the U.S. Department of Defense. The Pentagon labeled Anthropic a “supply-chain risk” — a designation historically reserved for foreign adversaries — after the company refused to let its AI be used for mass surveillance of Americans or autonomous weapons. Anthropic has filed for an emergency stay with the D.C. Circuit Court of Appeals, warning the label could cost it billions in lost revenue.

In a remarkable show of cross-industry solidarity, more than 30 OpenAI and Google DeepMind employees — including DeepMind chief scientist Jeff Dean — filed a statement supporting Anthropic’s position. Anthropic now wins roughly 70% of head-to-head enterprise matchups against OpenAI among first-time AI buyers, underscoring the business stakes.

NVIDIA GTC 2026: What to Expect Next Week

NVIDIA’s annual developer conference kicks off Monday, March 16, at the San Jose Convention Center, with CEO Jensen Huang delivering the keynote at 11 a.m. PT. The event is expected to showcase the full Rubin platform — six new chips delivering up to 5x inference performance and 3.5x training performance over Blackwell, with 288 GB of HBM4 memory and 22 TB/s of bandwidth.

Ahead of GTC, NVIDIA dropped Nemotron 3 Super, a 120-billion-parameter open model that activates only 12 billion parameters per forward pass. The hybrid Mamba-Transformer architecture supports 1-million-token context windows and is optimized for agentic reasoning. Full model weights, training data, and quantized checkpoints are all open-source.

AWS, Google Cloud, Microsoft Azure, and Oracle Cloud will be among the first to deploy Vera Rubin-based instances in the second half of 2026, alongside NVIDIA Cloud Partners CoreWeave, Lambda, Nebius, and Nscale.

Washington State Passes Landmark Chatbot Safety Bill

Washington lawmakers gave final passage to HB 2225, a first-of-its-kind AI companion chatbot safety bill that targets the growing wave of AI chatbot companions. The bill — the second such legislation to pass in 2026 after Oregon’s similar measure — requires chatbots to give users hourly reminders that they’re talking to an AI, not a real human.

The legislation mandates that developers implement suicidal ideation detection protocols, regularly report usage data, and prevent chatbots from showing explicit content to minors or simulating romantic relationships with them. Enforcement comes through Washington’s Consumer Protection Act, meaning any individual can sue for non-compliance. The bill was led by Rep. Lisa Callan and Sen. Lisa Wellman at the request of Gov. Bob Ferguson.

a16z’s Top 100 AI Apps: ChatGPT Hits 900M Weekly Users

Andreessen Horowitz released its latest ranking of the top 100 generative AI consumer apps, and the numbers are staggering. ChatGPT now has 900 million weekly active users, up 500 million over the past year and 2.5x larger than second-place Gemini on mobile. Claude, meanwhile, grew paid subscribers by over 200% year-over-year, while Gemini grew by 258%.

The report highlights the rise of AI agents as a consumer category, with OpenClaw becoming the most-starred project on GitHub (surpassing React and Linux) before being acquired by OpenAI. The a16z list now includes products where generative AI has become core to the experience, such as CapCut, Canva, Notion, and Grammarly.

Meta Delays ‘Avocado’ Model, Doubles Down on Custom Chips

Meta has postponed the release of its next-generation AI model, codenamed “Avocado,” to at least May 2026 after internal benchmarks showed it underperforming rivals. In the meantime, the company published technical details on its MTIA (Meta Training and Inference Accelerator) custom AI chip family, revealing four successive generations developed over two years to serve billions of users at lower cost.

Separately, Meta acquired Moltbook, the social network where AI agents post, comment, and vote on content. The Moltbook team is joining Meta’s Superintelligence Labs unit. The deal came after OpenAI hired the founder of OpenClaw, the agent framework that powered Moltbook.

Anthropic Launches The Anthropic Institute

Anthropic co-founder Jack Clark unveiled The Anthropic Institute, a new initiative aimed at advancing public debate on the challenges posed by powerful AI. The Institute will produce research, host convenings, and engage policymakers on AI safety, governance, and societal impact — arriving at a moment when the company’s own clash with the U.S. government has thrust these issues into the spotlight.