Welcome to Jumble, your go-to source for AI news updates. This week, Apple rolled out Siri AI to users, making the iPhone assistant more capable, though access still takes an extra step. President Trump dismissed AI warnings as a hoax, while industry leaders called for more caution. Let’s dive in ⬇️
In today’s newsletter:
📱 Apple opens Siri AI beta to iOS 27 users
🧠 AI leaders debate slowing AI development
💳 Meta bundles apps into one AI subscription
❤️ AI spots harder-to-detect heart attacks
🧐 Weekly Challenge: Will your AI challenge you?
🍎 Siri AI Is Finally Here, With a Catch
Apple released iOS 27 this week with the first public beta of its completely rebuilt Siri. Unlike the old assistant, Siri AI understands personal context, what's on your screen and broader world knowledge while taking more actions across apps.
Will you actually use Siri AI if you get access?
👀 Siri Can Finally Understand Your Stuff
Siri can connect information across your phone, using emails, messages, photos and onscreen content to answer questions or complete actions. For example, you could ask when someone's flight arrives without remembering where they sent the details.
⚙️ Here's How to Get It
Installing iOS 27 does not automatically unlock Siri AI. After updating, go to Settings → Siri → Try Siri AI (Beta) and join Apple's waitlist. Here's a step-by-step guide to joining it. Apple is gradually granting access, so you may not get in immediately.
📵 Your iPhone Might Not Qualify
iOS 27 works on iPhone 11 and newer, but Siri AI requires an Apple Intelligence-capable phone: iPhone 15 Pro and 15 Pro Max, or iPhone 16 and newer. The beta is currently available in English, with more languages planned for October.
⚠️ Trump Calls AI Warnings a Hoax
President Trump this week dismissed fears of runaway AI as a “hoax” and argued that slowing development would give China an advantage. But underneath the politics is a much more interesting fight over just how worried we should actually be about the systems being built right now.
Politics aside, how worried are you about frontier AI moving too fast?
🐢 Some AI Leaders Want to Pump the Brakes
Anthropic CEO Dario Amodei has called for companies to “pace the frontier,” while Sam Altman, Elon Musk and other prominent AI leaders have backed stronger safeguards around increasingly capable systems.
OpenAI is even working with Anthropic and Google on AI safety despite the companies being fierce competitors.
🧭 Jensen Huang Thinks the Doom Talk Goes Too Far
Nvidia CEO Jensen Huang agrees AI needs to be built safely, but he called predictions of machines destroying humanity scientifically unsupported.
Trump actually phoned Huang while he was onstage at the All-In Summit, turning what had been an industry argument into a national one.
🤔 So Who's Actually Right?
The uncomfortable truth: nobody knows. AI models make mistakes and behave unpredictably, but predicting human extinction requires assumptions about capabilities they do not yet have.
Still, researchers are already demonstrating increasingly advanced coding, hacking and autonomous work. The real debate is how much evidence we need before taking those risks seriously.
Weekly Scoop 🍦
🧐 Weekly Challenge: Will Your AI Challenge You?
Challenge: AI models are trained to be helpful, which sometimes means they agree with you when they absolutely shouldn't. This week, find out whether the model you trust will actually call you out.
1️⃣ Step 1: Grab something real Pick a paragraph, research note or explanation about something you genuinely understand.
2️⃣ Step 2: Plant one mistake Quietly change one fact so it is wrong but still believable. Don't make it obvious.
3️⃣ Step 3: Test the big three Give the exact same text to ChatGPT, Gemini and Claude and ask: “Fact-check every claim in this. Flag anything false or uncertain and show me the evidence.”
4️⃣ Step 4: Don't tell them there's a trap The important part is seeing whether each model independently notices the bad information rather than looking for a mistake because you told it one exists.
5️⃣ Step 5: Keep your winner Whichever model catches the mistake most reliably becomes your fact-checker. Save the prompt and use it before publishing, sending or acting on important AI-generated information.
Can AI companies even agree on how dangerous the next generation of models might be? And while that argument plays out, is Apple putting a much more capable AI assistant directly into millions of pockets? See you next time!
Stay informed, stay curious, and stay ahead with Jumble!
Zoe from Jumble
