Meta just put AI in the hands of 3.4 billion people for free. Here’s why that’s exactly the problem.
On April 8, Meta launched Muse Spark — the first model out of its new Superintelligence Labs, led by Alexandr Wang, who Zuckerberg poached from Scale AI for over $14 billion. It’s natively multimodal. It has a “Thinking” mode and an “Instant” mode. It counts calories from your photos. It answers health questions. And it’s rolling out across WhatsApp, Instagram, Facebook, Messenger, and smart glasses.
It scores 52 on the AI Intelligence Index — decent, but it trails GPT-5.4 (57), Gemini 3.1 Pro (57), and Claude Opus 4.6. That’s not the point. The point is: it’s free, it’s everywhere, and Meta is spending $135 billion this year alone to make sure it stays in your pocket.
💰 THE $135 BILLION QUESTION
Meta’s capital expenditure for 2026 sits between $115 and $135 billion — nearly double what it spent in 2025. Reality Labs hemorrhaged $4.2 billion last quarter. And 98% of Meta’s $200 billion in revenue still comes from advertising.
So who pays for the AI? You do. Just not with money. Not yet.
🪤 THE FREE TRAP
Here’s how this works. Muse Spark is embedded in apps you already open dozens of times a day. It learns your habits, your conversations, your health data, your shopping patterns. It becomes useful — then indispensable. Then Meta turns the knobs.
Three monetization lanes are already visible:
Subscriptions. A premium tier with deeper personalization, advanced Thinking mode, and health features. Not a feature inside the app — a standalone product. Sound familiar? It should. OpenAI just signalled the end of unlimited ChatGPT plans. Claude subscription prices may rise. The industry is moving from “AI for everyone” to “AI for everyone who can pay.”
Commerce in chat. AI shopping agents inside WhatsApp and Messenger that handle sizing, availability, shipping, and returns. Over 3 billion people use WhatsApp. In countries like Thailand and Vietnam, commerce already happens in chat. Meta isn’t building a store. It’s becoming the store.
Business agents. AI that answers customer questions, builds carts, and schedules appointments. This competes with customer support SaaS, not with Netflix. And it scales across every small business on Meta’s platform.
🔒 THE PROPRIETARY PIVOT
Here’s the part everyone should be talking about: Muse Spark is proprietary. Not open source. Not Llama.
When Llama 4 underwhelmed in April 2025, Zuckerberg spent billions overhauling Meta’s entire AI strategy. The result isn’t just a better model — it’s a closed one. You can’t monetize what you give away. Muse Spark is designed to make money, not friends.
This is a bigger shift than most people realize. Meta built its AI credibility on open source. Now that credibility has been cashed in for a walled garden with 3.4 billion users already inside it.
🏥 YOUR HEALTH DATA, META’S PRODUCT
Muse Spark’s health AI features — calorie counting from photos, health question answering — put it in direct competition with ChatGPT Health and Claude for Healthcare, both launched in January 2026. Health AI is the most sensitive category in consumer tech. Misinformation risk is real. Data privacy is non-negotiable.
And this is Meta. The company that turned personal data into the most profitable advertising machine in history. Now it wants your health data too.
🇪🇺 THE EU PROBLEM
Meta already faces EU regulatory pressure on its subscription models. European authorities flagged Meta’s “pay to avoid ads” option as non-compliant with the Digital Markets Act. Any AI monetization in Europe — especially involving health data or embedded commerce — will face intense scrutiny on consent, transparency, and data use.
🔗 THE BIGGER PICTURE
This isn’t just a Meta story. It’s the same pattern we’ve been tracking:
- Intelligence Withdrawal: Anthropic cut off third-party agent tools. Thinking depth dropped 67%. Same price, less intelligence. The subsidy era is ending everywhere.
- AI Price Divide: As frontier models get more expensive, the rich get Claude Mythos while everyone else gets last year’s tech.
Meta’s version is just smoother. It doesn’t charge you directly. It embeds AI into 3.4 billion daily interactions and makes you pay with attention, data, and eventually commerce. The product isn’t the AI. The product is you.
⚡ WHAT TO WATCH
- How quickly Meta introduces a paid tier for Muse Spark
- Whether the proprietary model holds or faces developer backlash
- EU regulatory response to health AI features and embedded commerce
- Whether OpenAI and Anthropic raise consumer prices in response
- If thinking depth and model quality degrade over time (the shrinkflation pattern)
Free AI isn’t free. It never was. Muse Spark just makes the bill harder to see.
🔍 THE BOTTOM LINE
This isn’t new. Big tech has always made you the product:
- Google gives you search, Gmail, Maps, and YouTube for free — and makes $307 billion a year from ads built on your data, your clicks, your location history.
- Facebook (before it was Meta) gave you a social network for free — and sold access to your attention to every advertiser on earth.
- TikTok gives you endless entertainment for free — and its algorithm learns more about your psychology in a week than your therapist learns in a year.
- Amazon gives you Prime video and free shipping — and locks you into an ecosystem where it knows what you buy, when you buy it, and what you’ll buy next.
- Apple gives you iMessage and iCloud — and keeps you trapped in a $1,200 phone upgrade cycle because green bubbles are social death.
Meta’s playbook with Muse Spark is the same story with a new character. Give it away. Get billions hooked. Harvest the data. Monetize the dependence. The difference now? AI doesn’t just know what you like — it knows how you think. It answers your health questions. It shops for you. It thinks for you.
The cost of free AI isn’t money. It’s autonomy. And by the time the subscription screen appears, you’ll already be too dependent to click cancel.
