Pentagon, Chatbots, Data Collide
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Saturday, March 7, 2026
Three stories this Saturday, three different actors, one shared question: who actually gets to decide what AI can and can't do?
The Pentagon is punishing Anthropic for saying no, New York lawmakers want to stop chatbots from playing doctor, and a federal judge just told Elon Musk that California can peek at his training data. The people building AI and the people writing the rules are on a collision course, and this week, the collisions got loud.
Today in AI:
- Anthropic Gets the Cold Shoulder from the Pentagon - The Department of Defense officially designated Anthropic a supply-chain risk after the company refused to grant unrestricted access to Claude for mass surveillance and autonomous weapons. Anthropic plans to fight the designation in court, while Microsoft, Google, and AWS confirmed Claude remains available to non-defense customers. TechCrunch
- OpenAI Swoops In on the Pentagon Deal, Then Faces the Fallout - OpenAI signed a classified-work contract with the Defense Department the same day Anthropic's deal collapsed. CEO Sam Altman acknowledged the timing was a bad look, and ChatGPT reportedly saw a 295% surge in uninstalls from users unhappy with the move. Fast Company
- New York Wants AI Chatbots to Stop Playing Doctor and Lawyer - A New York bill passed committee 6-0 and would ban AI chatbots from dispensing medical or legal advice, covering 15 licensed professions. Uniquely, it gives harmed users the right to sue, even if the chatbot clearly labels itself as AI. Fast Company
- Musk Loses Bid to Block California's AI Data Disclosure Law - A federal judge ruled that xAI failed to show its training datasets qualify as trade secrets, denying a request to halt a California law requiring AI companies to disclose their data sources. The judge also dismissed xAI's First Amendment claims. Ars Technica
- Claude Found 22 Firefox Vulnerabilities in Two Weeks - Anthropic partnered with Mozilla and used Claude Opus 4.6 to discover 22 vulnerabilities in Firefox, 14 of them high-severity. That's nearly a fifth of all high-severity Firefox bugs remediated in 2025, found in a fraction of the time. Anthropic
- Americans Want AI Guardrails More Than They Want to Beat China - A new survey from the Human Artistry Campaign found that support for AI guardrails and copyright protections is more bipartisan than believing the US should lead China in tech. Translation: voters care more about controlling AI than winning the AI race. Semafor
- Google Opens Up Speech Tech for 27 African Languages - Google Research released WAXAL, an open dataset covering 27 Sub-Saharan African languages with over 1,800 hours of transcribed speech and 565 hours of text-to-speech recordings. The goal is to bring voice-enabled tech to more than 100 million speakers currently left out. Google Research
- This $1,199 Orb Wants to Jam AI Wearable Microphones - A Harvard grad's startup Deveillance announced Spectre I, a portable device designed to block always-listening AI wearables from recording your voice. The catch: physicists and engineers are deeply skeptical it can actually deliver on that promise. Wired
Today's Takeaway:
The Anthropic-Pentagon clash isn't just a contract dispute - it's exposing how little framework exists for what happens when the government wants AI tools and the company that built them says "not like that." Think of it like a landlord-tenant standoff, except the landlord is the Defense Department and the tenant built the house. The supply-chain risk designation, a label typically reserved for foreign adversaries, is being used here as a pressure tactic against an American company. As Zvi Mowshowitz argues, less disruptive legal tools were available, and talks were reportedly close to a deal before a leaked internal memo from CEO Dario Amodei inflamed the situation further.
Meanwhile, MIT Technology Review points out the legal vacuum at the center of this fight: existing surveillance law was written before the internet existed, let alone AI that can aggregate innocuous data into detailed profiles at scale. OpenAI stepped in to fill Anthropic's spot, but the reported 295% surge in ChatGPT uninstalls, per TechCrunch, suggests users noticed. Here's the thing: this fight isn't really about one contract. It's setting the precedent for whether AI companies can say no to the most powerful customer in the world.
💡 Fluency Moment - Building your AI fluency, one term at a time.
"Training Data"
In plain English: The massive collection of text and information used to teach an AI how to respond.
Think of it like: Like a student's entire reading history - every book shapes how they think and answer questions.
Why you'll hear about it: Courts are now forcing AI companies to reveal what data trained their models.
🧰 Your Toolkit
- Your AI Security Checklist: Staying Safe in a World of AI and Digital Threats - [ ] Check if any app or tool you use mentions 'AI-powered' features - know what data it collects before you sign up.
- Never type your home address, phone number, or passwords into a free AI chatbot like ChatGPT or Gemini.
- If an AI tool asks for access to your email or files, ask yourself: does it really need that to do its job?
- Search '[App Name] data privacy' before trusting any new AI tool with personal information.
- Turn on two-step login (also called two-factor authentication) for any account that uses AI features.
- If you use AI at work, ask your manager whether your company has rules about what information you can share with AI tools.
- Once a month, review which apps on your phone have permission to use your microphone or camera - remove ones you don't recognise.
These simple checks help you stay in control as AI becomes part of everyday life. A few minutes of caution today can protect your privacy for years to come.
The Bottom Line
The Pattern: Across this Saturday's stories - Anthropic vs. the Pentagon, New York's chatbot bill, xAI's courtroom loss, even a $1,199 anti-surveillance orb - every fight is about the same thing: who controls what AI does once it's out in the world. The builders, the regulators, the military, or you?
Why It Matters: The rules being written right now, in courtrooms, committee votes, and Pentagon memos, will determine whether AI companies can set boundaries on their own products or whether customers (especially government ones) get to dictate terms. That precedent affects every AI tool you'll use for the next decade.
Your Move: Read Anthropic's actual usage-vs-capability data from their labor market report - the gap between what AI theoretically can do (94% of computer science tasks) and what it actually does (33%) is the most honest reality check the industry has produced. Use it the next time someone tells you AI is about to replace everything overnight.
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