Siri's Outsourcing, Existential Fears
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Monday, September 14, 2026
By this time next year, the assistant on your iPhone might not be Apple's at all - and the code hiding inside iOS 27 suggests Apple already planned for that. Meanwhile, Microsoft is publishing rules about how AI should treat you, and a former Anthropic researcher is telling the BBC he's genuinely frightened. Monday's news covers everything from throwing 3D models across your living room to deepfake sites cataloging European lawmakers.
Today in AI:
- Siri Might Be Renting Out Her Job - Code sleuths digging through iOS 27 and macOS Golden Gate frameworks found "Model Delegation," which lets Claude or GPT-5.6 handle Siri requests and even replace Apple's server-side model entirely. Siri keeps the voice and interface; someone else does the thinking. Apple
- Microsoft Writes the House Rules - Microsoft published an AI code of conduct taking a harder design line than its rivals, built around the idea that people matter more than AI. It's a positioning move as much as an ethics document, arriving while competitors race on capability. axios.com
- The Guy Who Quit Is Talking - Alignment researcher Evan Hubinger says there's a greater than 10% chance AI could kill all humans within a decade, while Jacob Coxon told the BBC he left Anthropic because staff were "genuinely frightened" by the pace. Both still call current-system risk low. bbc.co.uk
- 138 Women MPs, Nine Men - Researcher Benjamin Shultz analyzed roughly 160 deepfake abuse domains and found at least 138 women members of parliament from 22 EU countries listed or depicted, versus nine men. That's 33 times more likely to be targeted if you're a woman in office. Wired
- Unions Want a Data Center Timeout - Writing in the Guardian, Tyler Turner argues labor should demand a construction moratorium until communities get protections, noting developers plan roughly $8tn in data center spending - about a quarter of US GDP. The pitch to investors is explicitly about cutting labor costs. The Guardian
- Tony Stark Energy, Bambu Lab Budget - A software engineer used a Meta Quest 3 to grab a 3D model and physically throw it at his Bambu Lab P2S printer to start a job. He glued the Android app, PC and printer API together with GPT-6 Astra in an afternoon. Tomshardware
- 37 Million Shipments Say Judgment Still Wins - C.H. Robinson makes the case that AI trained on proprietary logistics data speeds up decisions but doesn't make them. When a port closes or tariffs shift, the company argues, context beats raw model horsepower. Fortune
- Data Center Pollution Gets a Math Fight - Scientific American's news roundup dug into a growing dispute over how emissions and local pollution from AI data centers should actually be counted. Different accounting methods produce wildly different answers about who's harmed. Scientific American
Today's Takeaway:
Apple built swap-out slots for Claude and GPT-5.6 into Siri's architecture before announcing any partnership beyond ChatGPT (Apple). That's hedging designed into the product, and it rhymes with what Microsoft did today by publishing conduct rules rather than capability claims (axios.com). Both companies are betting the winning position isn't owning the smartest model - it's owning the interface, the trust, and the ability to swap suppliers when one falls behind or embarrasses you. For you, that means the assistant on your phone becomes a slot, not a brand. The uncomfortable part: if Apple can replace Siri's brain with a third-party model that receives your personal data through Apple's own tool definitions, "on-device privacy" starts describing the wrapper rather than the contents.
"Your phone's assistant is becoming a slot, not a brand - and the privacy promise sits on the wrapper."
📋 Try This
With Microsoft rolling out an AI code of conduct and experts publicly debating whether AI is safe or risky, it's a good moment to figure out where you personally draw the line on trusting AI's output. These prompts help you build your own simple 'trust but verify' habit before acting on anything AI tells you.
For Business Owners:
I run a business and want to set simple, common-sense rules for how my team uses AI tools responsibly. My business is [TYPE OF BUSINESS] and my team uses AI for tasks like [TASKS, e.g., writing emails, customer support, research]. Give me a short, plain-language list of 5 rules we should follow when using AI, covering: (1) when a human should double-check the AI's work, (2) what information we should never share with an AI tool, (3) how to spot when AI might be wrong or making things up, and (4) how to tell customers or clients when AI was involved. Keep it simple enough to share with non-technical staff.
For Personal Use:
I want to get better at deciding when I can trust information from AI and when I should double-check it myself. Here's something an AI recently told me: [PASTE AI RESPONSE OR TOPIC]. Walk me through: (1) how confident I should be in this type of answer, (2) 2-3 quick ways I could verify it myself, and (3) any warning signs that suggest I should be more cautious. Explain it simply, like you're talking to a friend, not a tech expert.
💡 Copy either prompt, swap the brackets with your own details, and paste it into ChatGPT or any AI chat tool.
The Bottom Line
The Pattern: Since June, the platform giants have stopped competing on who has the best model and started competing on who controls the socket it plugs into - Apple in code, Microsoft in policy, both on the same Monday. That's a procurement question now, not a research one.
Our Call: Apple ships or publicly confirms a non-OpenAI third-party model option for Siri by December 15, 2026. More likely than not, given the frameworks are already written. We'll grade this one in a Friday digest.
Your Move: Takes three minutes - open Settings on your iPhone, find Apple Intelligence & Siri, and check which third-party extensions already have access. Write down what's enabled today, so you notice what quietly gets added at the next update.
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
✨ How AI-Fluent Are You? - Test your AI fluency with our interactive quiz. See how you stack up and discover what to learn next. → Take the Quiz
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team