Weekly AI News Roundup — March 2026

Another week, another wave of AI developments. This week's biggest stories span from big tech's latest AI pushes to a legal battle between Anthropic and the US government.

Nvidia Planning Open-Source AI Agent Platform

Nvidia is reportedly planning to launch an open-source platform for AI agents, according to sources familiar with the company's plans WIRED. This marks a significant move for Nvidia, which has been primarily known for its hardware dominance in the AI space. The platform would allow developers to build and deploy AI agents more easily, potentially democratizing access to agentic AI technology.

Additionally, Nvidia announced a massive billion investment in photonics technology, splitting billion each between Lumentum and Coherent to develop data center networking technology that could dramatically speed up AI model training The Verge.

Anthropic Sues the Pentagon

In a bold move, Anthropic has filed a lawsuit against the Pentagon, with employees from OpenAI and Google filing an amicus brief in support MIT Technology Review. The lawsuit challenges the Pentagon's AI contract restrictions, which Anthropic argues could threaten military missions. Meanwhile, OpenAI is reportedly looking at a contract with NATO, stepping into the void left by Anthropic's refusal Reuters.

Google Goes All-In on Gemini

Google continues to push Gemini aggressively across its product suite. This week saw the rollout of new Gemini capabilities to Google Docs, Sheets, Slides, and Drive TechCrunch. The company also announced Gemini 3.1 Pro, claiming improvements in complex problem-solving. Separately, Google's AI Mode with Canvas is now rolling out to all US users TechCrunch.

Zoom Enters the AI Office Suite Race

Zoom introduced an AI-powered office suite, signaling its ambition to compete with Microsoft and Google in the productivity space TechCrunch. Perhaps most notably, Zoom announced that AI avatars for meetings will arrive this month—allowing users to send digital representatives to meetings in their place.

YouTube Expands Deepfake Detection

YouTube is expanding its AI deepfake detection tool to cover politicians, government officials, and journalists TechCrunch. This comes as concerns about AI-generated content ahead of elections continue to grow.

LLMs Can Unmask Pseudonymous Users

Researchers have found that LLMs can identify pseudonymous social media accounts with surprising accuracy Ars Technica. This raises serious privacy concerns for users who rely on anonymity online.

OpenAI's 'Best Model Ever'

OpenAI has released what it calls its "best model ever," though details are still emerging The Rundown AI. Meanwhile, CEO Sam Altman faces fallout from OpenAI's Pentagon deal decisions The Rundown AI.

US Mulls New AI Chip Export Rules

The US government is considering new rules for AI chip exports, including requiring US investments by foreign firms Reuters. This adds to the growing tensions around global AI hardware distribution.

Stay tuned for more AI news next week. This roundup is brought to you by Conclio.