AI News Summary - Week 12.2, 2026

Another action-packed week in AI. OpenAI continues its aggressive model rollout while the broader tech industry unites behind Anthropic in an unprecedented legal showdown with the Pentagon.

OpenAI Releases GPT-5.4 mini

Building on the GPT-5.4 launch from last week, OpenAI rolled out GPT-5.4 mini on March 18, making the more efficient model available to free users through ChatGPT's "Thinking" feature. The full GPT-5.4, released March 5, introduced native computer use capabilities—allowing the model to interact directly with computers, a significant step toward truly autonomous AI agents.

The Verge noted that GPT-5.4 represents "a big step toward autonomous agents," with the ability to execute tasks on users' computers rather than just generating text. This positions OpenAI squarely in the race to build AI that can take action, not just reason.

Tech Industry Unites Behind Anthropic vs. Pentagon

In a remarkable display of industry solidarity, employees from OpenAI and Google filed an amicus brief supporting Anthropic in its lawsuit against the Pentagon. The case centers on a February 2026 federal ban preventing Anthropic from entering defense contracts.

As The Verge reported, more than 30 employees from OpenAI and Google signed the brief, arguing that the ban sets a dangerous precedent for AI development and could stifle critical safety research. This marks a rare moment where direct competitors have publicly aligned against government overreach into AI.

The lawsuit could have far-reaching implications for how AI companies interact with the defense sector and how the government regulates advanced AI systems.

Nvidia Doubles Down on AI Agents

Nvidia announced its Open Physical AI Data Factory Blueprint on March 16, a new framework designed to accelerate robotics, vision AI agents, and autonomous vehicle development. This reinforces CEO Jensen Huang's vision of AI agents "being everywhere"—a message he's been pounding at every opportunity.

The blueprint provides hardware and software infrastructure for companies building physical AI systems, signaling Nvidia's intent to own the stack from chips to robotics platforms.

Quick Hits

• Claude Opus 4.6: Anthropic upgraded its smartest model in early February with improved coding skills and better agentic task performance
• Gemini 3.1: Google continues pushing its Gemini series, with Flash-Lite updated in early March

Sources: TechCrunch, The Verge, Fortune, CNN, NVIDIA Newsroom