1. White House Unveils National AI Policy Framework, Moves to Preempt State Laws
The Trump administration released a six-pronged national AI policy framework on March 20, seeking to federalize AI regulation and limit state-level oversight. The framework covers child safety rules, data center permitting and energy standards, and establishes an AI Litigation Task Force to challenge state laws deemed "onerous." States with restrictive AI regulations could lose eligibility for federal broadband funding under the new guidance.
2. Three Charged with Smuggling Nvidia AI Chips to China
The U.S. Department of Justice charged a senior vice president of Super Micro Computer and two associates with conspiring to illegally export billions of dollars worth of servers containing advanced Nvidia AI chips to China. The case highlights the intensifying battle over AI hardware supply chains and export controls as the U.S. tightens restrictions on cutting-edge semiconductor technology reaching adversaries.
3. Meta Launches Llama 4 with 10-Million-Token Context Window
Meta released Llama 4, its next-generation open-source model family built on a Mixture-of-Experts architecture. The lineup includes Llama 4 Scout, which boasts an industry-leading 10-million-token context window, and Llama 4 Maverick, both natively multimodal models capable of processing text, images, and short video. The massive context window makes Scout particularly suited for enterprise document analysis at scale.
4. Anthropic Ships Claude Opus 4.6 and Sonnet 4.6 with Extended Context
Anthropic launched Claude Sonnet 4.6 and Opus 4.6, focusing on code generation, reasoning, and extended context capabilities. Sonnet 4.6 introduces a 1-million-token context window in beta, while Opus 4.6 features adaptive thinking that dynamically adjusts reasoning depth based on task complexity. Independent evaluator METR measured Opus 4.6's task-completion horizon at approximately 14.5 hours, the highest recorded for any model.
5. Elon Musk Announces SpaceX-xAI Merger
Elon Musk revealed plans to merge SpaceX with his AI venture xAI, aiming to deeply embed xAI's Grok models into SpaceX operations including satellite constellation management and mission planning. Morgan Stanley analysts warned the move signals an "AI leap" in 2026 that most industries aren't prepared for, as AI becomes tightly integrated with physical infrastructure and real-time decision systems.
6. NVIDIA Launches Agent Toolkit for Autonomous Enterprise AI
NVIDIA released its Agent Toolkit, an open platform enabling developers to build autonomous AI agents capable of multi-step reasoning, tool use, and complex enterprise task completion. The toolkit arrives alongside Google's rollout of its Personal Intelligence feature to all U.S. users, allowing Gemini to draw on connected app data from Gmail, Photos, and YouTube for free-tier users.
7. Yann LeCun Raises $1B+ for Open-Source 'World Models' Startup AMI
Yann LeCun, Meta's former chief AI scientist, secured over $1 billion in funding for his startup AMI, which is building open-source "world models" designed to understand and reason about physical environments. The company aims to create AI systems that move beyond text prediction toward genuine understanding of how the world works, a goal LeCun has long advocated as essential for artificial general intelligence.
8. Atlassian Cuts 1,600 Jobs to Redirect Resources Toward AI
Atlassian announced it is laying off roughly 10% of its global workforce — approximately 1,600 employees — to redirect resources toward AI development and enterprise sales. The restructuring reflects a broader industry pattern where established tech companies are aggressively reallocating budgets from traditional product lines to AI capabilities, betting that AI-native features will determine competitive positioning in the enterprise software market.
// KEY TAKEAWAYS
This week marks a pivotal moment in AI governance as the White House moves to centralize AI regulation at the federal level while chip export enforcement intensifies with criminal charges against Supermicro executives. On the model front, the context window race has exploded — Meta's Llama 4 Scout pushes to 10 million tokens while Anthropic and OpenAI compete at the 1-million-token tier. Meanwhile, the blurring of AI with physical infrastructure (SpaceX-xAI merger) and massive open-source bets (LeCun's $1B+ AMI) signal that 2026 is the year AI stops being a software story and becomes an everything story.