Who Owns You Online? AI Clones, Shopping Bots, and the Fight to Keep the Web Human
Three stories today share a quiet thread: AI is increasingly standing between you and the internet — deciding what you buy, who speaks for you, and what gets through. That shift is happening faster than most people realize, and the people building the infrastructure are starting to ask hard questions about it.
Cloudflare’s CEO Thinks the Web Needs Defending — From AI
Matthew Prince runs Cloudflare, the company that quietly handles security and traffic for a huge chunk of the internet. In a new interview with The Verge, he reportedly talked through how his company is thinking about the damage AI could do to the web — and what Cloudflare might do about it.
The concern isn’t science fiction. AI systems can flood websites with fake traffic, scrape content without permission, generate convincing misinformation, and make it nearly impossible to tell a human from a bot. Cloudflare sits in a unique position here: almost every website that uses its services passes through Cloudflare’s infrastructure first, like traffic through a tollbooth. That gives the company unusual visibility into what’s actually happening on the web.
For everyday people, this matters because the health of the open web affects everything from the news you read to the small businesses you shop from. If AI-generated content and bot traffic crowd out real human activity, the internet becomes a less useful and less trustworthy place. Prince’s argument, according to The Verge, is that companies like Cloudflare may be among the few with enough reach to actually do something about it.
Why this matters: The web’s problems with AI aren’t just a tech industry headache — they affect anyone who relies on the internet for information, shopping, or communication. Cloudflare’s next moves could shape what the web looks like for years.
Matthew Prince, Cloudflare CEO, on protecting the internet from AI harms.
Google Is Testing AI Shopping in India — and It’s a Big Hint at the Future
According to TechCrunch, Google is reportedly piloting a feature that lets users in India buy products from Flipkart — the Walmart-owned e-commerce giant — directly through Gemini, Google’s AI assistant. A small group of users can already try it, with a broader rollout expected before the end of the month.
This is what people in tech call “agentic” shopping — where an AI doesn’t just give you information but actually takes actions on your behalf, like completing a purchase. Think of it less like a search engine and more like a personal shopper who can check out for you. Instead of opening the Flipkart app, finding a product, and entering your payment details, you’d just tell Gemini what you want.
For shoppers, the appeal is obvious: less clicking, faster buying. But this arrangement also raises questions worth sitting with. When an AI chooses which products to show you, whose interests is it serving? The AI company’s, the retailer’s, or yours? India is often a testing ground for new tech features before they roll out globally, so what works there tends to show up everywhere else soon after.
Why this matters: If AI assistants become the primary way people shop online, the company running that assistant gains enormous power over what gets bought and what gets buried.
Shopping on Flipkart through Gemini rather than the website or app directly.
What Happens When Anyone Can Build a Digital Copy of Themselves?
A writer at TechCrunch reportedly built a conversational AI avatar — a digital version of themselves that can hold real-time conversations — trained on their own writing, voice, and personal data. The experiment worked. And that’s exactly what made it unsettling.
Building this kind of avatar is now accessible enough that a journalist, not an engineer, can do it over a weekend. You feed the system your emails, social posts, or recordings, and it learns to mimic how you think and talk. The result isn’t a simple chatbot — it’s something that sounds like you, reasons somewhat like you, and can speak to people even when you’re not there.
The practical implications are already arriving. Celebrities could offer fans a version of themselves that’s always available. Executives could have an AI stand-in for routine meetings. Grieving families could, in theory, talk to a version of someone who has died. None of these are hypothetical anymore. The TechCrunch piece focuses on a simpler but equally important question: if someone builds a fake version of you, what can you do about it?
Why this matters: The technology to clone someone’s conversational identity is now cheap and accessible. Society’s rules for when that’s acceptable — and when it isn’t — are nowhere near ready.
Also Happening in AI
On the developer tools front, Anthropic quietly released version 1.8.0 of its Python SDK — a coding library that developers use to build apps with Claude — adding support for the new Claude Opus 5.5 model. Microsoft’s AutoGen library, used for building networks of AI agents that work together, pushed out version 0.7.5 with documentation fixes. LlamaIndex, another popular open-source toolkit for AI app development, hit v0.14.25. AutoGPT released platform beta v0.8.1, reportedly adding eight new AI specialist agents. On the infrastructure side, TechCrunch reports that Crusoe Energy has walked away from a $1.25 billion plan to power AI data centers using experimental stationary turbines built by Boom Supersonic — a reminder that the race to find cheap, reliable power for AI is still very much unsettled.
What to Watch
The Flipkart-Gemini pilot is the one to track closely: if Google expands it to other markets and other retailers, it signals a fundamental shift in how e-commerce works — away from open search results and toward curated AI recommendations. Watch also for how Cloudflare articulates a public policy position on AI traffic. If a major infrastructure company starts making editorial decisions about which AI systems get access to the web, that’s a form of power with almost no regulatory framework around it yet.