AI Weekly News Summary - Week 10, 2026

Another busy week in the AI world. Here's what's been happening.

OpenAI Steps Into Anthropic's Pentagon Void

The biggest story this week: OpenAI has signed an agreement with the US Department of Defense, filling the void left by Anthropic after their contract was terminated.

The Pentagon had labeled Anthropic a "supply chain risk" after the company refused to give unrestricted access to its Claude models for autonomous weapons and mass surveillance. Defense Secretary Pete Hegseth called Anthropic's approach "arrogance and betrayal."

OpenAI's deal takes a different approach—rather than specific prohibitions, the company is relying on existing laws to restrict how the military can use its technology. CEO Sam Altman added that Anthropic "seemed more focused on specific prohibitions in the contract, rather than citing applicable laws."

The key tension: OpenAI says it will maintain control over safety rules and won't give the military a version of its AI stripped of safety controls. But critics argue this is a compromise that doesn't go far enough. MIT Technology Review

Google Releases Nano Banana 2

Google has rolled out the latest version of its AI image generator. The update brings improved quality and faster generation times. Nano Banana 2 continues Google's push in the image generation space, competing with tools like DALL-E and Midjourney. Wired

Meta Tests AI Shopping Tool

Meta is rolling out a new AI-powered shopping tool to testers in the US, aiming to compete with similar offerings from ChatGPT and Gemini. The tool helps users find products and make purchasing decisions through conversational AI. The Verge

Perplexity Signals Strategic Shift Away From Ads

Perplexity is retreating from its advertising-focused strategy, signaling a bigger shift in how AI search companies plan to monetize their products. The move raises questions about the future of AI-powered search monetization. Wired

The Rundown AI Highlights

• Supreme Court ducks AI copyright question — The court declined to rule on whether AI-generated content can be copyrighted, leaving the legal landscape unclear.
• Local AI transcription — New tools allow free video transcription entirely offline using local AI models.
• Perplexity's 19-model AI 'Computer' — A new approach to AI that combines 19 different models. The Rundown AI

Other Notable Stories

• Iowa data center zoning — An Iowa county adopted strict zoning rules for data centers, some of the toughest in the US, as communities push back against AI infrastructure growth. Ars Technica
• LLMs can unmask pseudonymous users — New research shows LLMs can identify anonymous users with surprising accuracy, raising privacy concerns. Ars Technica
• GrapheneOS and Motorola partnership — Could lead to more secure smartphone options beyond Pixel devices. The Verge

That's it for this week! Stay tuned for more AI updates next Wednesday.

News sources: TechCrunch, MIT Technology Review, Wired, The Verge, Ars Technica, The Rundown AI