AI Agents: War, Phones, Failures
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Sunday, March 1, 2026
Three stories this Sunday, three threads of the same rope: Anthropic gets banned from the Pentagon and promptly rockets to the No. 1 app in America, Honor straps a robotic arm onto a smartphone, and a CNBC deep dive warns that AI failures are spreading silently through businesses before anyone notices. The pattern?
AI agents are pushing into every corner of life - military ops, your pocket, your company's back office - and the guardrails are still being welded on mid-flight.
Today in AI:
- Anthropic Gets the Boot, Then Gets the Downloads - The Trump administration moved to ban Anthropic's Claude from federal agencies after a contract dispute over military use restrictions. Within 24 hours, Claude shot to No. 1 on Apple's free app chart. Getting rejected by the Pentagon turns out to be great marketing. Cnbc
- Claude Helped Bomb Iran Hours After Getting Banned - According to the Wall Street Journal, the U.S. military used Anthropic's AI tools during a major air strike on Iran just hours after Trump ordered agencies to stop using the company's tech. A six-month phase-out period means Claude is still in the warfighter's toolkit - for now. Engadget
- Palantir: The Quiet Giant Behind Pentagon AI - While Anthropic and the Defense Department argue publicly, Palantir quietly holds the keys. The company controls the classified data infrastructure that any replacement AI model will need to plug into. Translation: switching from Claude isn't as simple as downloading a new app. Fast Company
- Honor Puts a Robot on Your Phone (Literally) - At MWC in Barcelona, Honor unveiled a phone with a 200-megapixel camera mounted on a pop-up robotic gimbal arm. It tracks objects, nods yes or no, and bobs to music. It's weird, it's charming, and it's Honor's bet on standing out in a sea of identical slabs. Engadget
- 'Silent Failure at Scale' Is the AI Risk Nobody Sees Coming - A CNBC report highlights a growing problem: AI systems deployed in real business operations are making small, quiet mistakes that compound over weeks into major operational damage. The AI model builders themselves admit they don't know where the technology is headed. Cnbc
- Alibaba Open-Sources CoPaw, an Agent Workstation - Alibaba researchers released CoPaw, a free framework for building AI agents that can juggle multiple workflows and maintain memory across sessions. It's aimed at developers who want to move beyond simple chatbot interactions into persistent, multi-channel AI assistants. MarkTechPost
- UK Data Centres Face Climate Pressure - Campaign groups are pushing the UK government to force data centre developers to disclose their net emissions impact, warning that AI infrastructure could double national electricity demand. In Scotland, one proposed 250-acre facility is meeting fierce local resistance. The Guardian
- Anthropic Launches a 'Memory Import' Prompt - Anthropic quietly shipped a feature at claude.com/import-memory that lets users export all stored preferences, personal details, and instructions from another AI service and bring them into Claude. It's a customer-acquisition play dressed as a data portability tool, and it's clever. Simon Willison
Today's Takeaway:
The Anthropic-Pentagon saga is the most consequential AI story this Sunday, and it's tempting to read it as a simple political spat. It's not. Here's the thing: Anthropic refused to let the military use Claude for mass surveillance or fully autonomous weapons. The Defense Department demanded unrestricted access for "all lawful purposes." When neither side blinked, Trump publicly banned the company from government contracts. Yet as Engadget reported, the military used Claude in an air strike on Iran that same day. The six-month phase-out means Anthropic's tech is still embedded in active operations, and as Fast Company detailed, replacing it requires navigating Palantir's classified data infrastructure - a process that could take months longer than anyone admits.
The twist nobody expected: consumers rallied. Cnbc tracked Claude surging past ChatGPT to No. 1 on Apple's App Store by Saturday. The memory-import tool reported by Simon Willison is no coincidence - Anthropic is capitalizing on the spotlight to pull users from competitors while public sympathy runs high. Let's be real: refusing the Pentagon turned out to be the most effective product launch campaign of the year.
💡 Fluency Moment - Building your AI fluency, one term at a time.
"AI Agent"
In plain English: An AI that takes actions on your behalf, not just answers questions.
Think of it like: A virtual intern who doesn't just give advice but actually does the task for you.
Why you'll hear about it: AI agents are now handling military ops, phones, and business back offices.
🧰 Your Toolkit
5-Minute Quickstart: How to Use AI Tools Safely as a Complete Beginner
- Open ChatGPT or Google Gemini in your browser and create a free account using your email address.
- Type a simple question you'd normally Google, like 'What are easy dinner ideas for a busy weeknight?' and read the response.
- Ask a follow-up question like 'Make it vegetarian' to see how AI remembers your conversation like a helpful assistant.
- Test AI awareness by asking 'What are the risks of trusting AI answers without checking them?' - notice it admits its own limits.
- Try a real task: type 'Summarize this in simple terms: [PASTE ANY NEWS HEADLINE OR PARAGRAPH HERE]' and see the result.
- Before acting on any AI response, spend 30 seconds checking one fact it told you using a trusted website like a news outlet or official source.
Once you're comfortable asking questions, try using AI to help you draft a short email or plan your week. The more you practice with low-stakes tasks, the more confident you'll feel using it for bigger ones.
The Bottom Line
The Pattern: Every major story this Sunday - the Pentagon ban, silent AI failures in business, a phone growing a robotic limb, open-source agent frameworks - points to the same thing: AI is no longer a tool you choose to use. It's infrastructure that's being woven into military ops, consumer devices, and corporate workflows simultaneously, and the people building it openly admit they can't predict where it's going.
Why It Matters: The Anthropic saga shows that the question of who controls AI deployment is now a live political and commercial battleground, not an academic debate. Meanwhile, the CNBC report on silent failures means companies adopting AI agents without monitoring are accumulating invisible risk - the kind that shows up as a compliance crisis six months from now, not a flashing red alert today.
Your Move: If your team is using any AI tool in a workflow that touches customers, finances, or decisions, set up a weekly spot-check this week. Pull five random AI outputs and verify them against the source data. That 30-minute habit is the cheapest insurance against the "silent failure at scale" that's already hitting companies who assumed AI would just work.
What We're Working On
✨ Founding Cohort Special - 60% Off! - Use code MAF20 to join for just $20/month (regularly $50). Get weekly group sessions & workshops, self-paced courses for all levels, access to tools & templates, challenges with peer feedback, and 24/7 support community. → Join Now
✨ Free 30-Minute AI Consultation - Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team