Charlie Makes an Appearance: Inrupt at SXSW London
Inrupt’s primetime session at SXSW London last week started with a question.
"Who here doesn't quite trust AI completely?" asked CNBC correspondent Tania Bryer from the stage.
Nearly every hand in the room went up.
The Problem with Platform AI
The show of hands was striking, but it wasn't surprising. The data has been pointing this way for some time.
Pew Research found that 50% of Americans are more concerned than excited about AI in daily life — up from 37% in 2021. Edelman found that nearly three times as many Americans actively reject the growing use of AI (49%) as embrace it (17%).
It’s not so much backlash against the tech itself, but against who controls it.
"People are worried it'll go the same way as the web," Tim explained on stage. The web began as something genuinely pro-human: open, decentralized, for everyone. Over time, enormous businesses emerged by harvesting just enough data about people to sell targeted advertising. It came to be known as the Attention Economy.
But the AI era introduces something more intimate, and more consequential. At Inrupt we call it the “Memory Economy.”
"The Attention Economy was one degree of web we had," John explained. "Way worse is this memory notion." Unlike social platforms that learned to predict what we'd click, LLMs accumulate a comprehensive, continuous understanding of who we are — our preferences, our finances, our health, our relationships. They never forget. And if that memory is owned by the platform rather than the individual, it’s a situation ripe for misuse and manipulation.
Businesses At Risk, Too
John was equally direct about the stakes for businesses watching this moment from the sidelines. Platform AI — AI that centralizes data, infers preferences, and builds loyalty to itself rather than to the company selling product and services through it — will supplant the brands and institutions that rely on customer relationships to survive. "They'll all be disintermediated," he said plainly.
The alternative isn't a retreat from AI. It's a different architecture for it — one where the customer's data stays with the customer, where consent is real and revocable, and where the value of a personal AI relationship flows back to the company that the customer actually chose.
"Organizations have a chance here to do the right thing," John said. And create opportunity in doing it.
How Charlie Changes the Equation
The core infrastructure Inrupt has spent years building is a data vault: a secure, user-controlled store for the full universe of personal data. Built on Tim's Solid protocol and managed through Inrupt's Enterprise Solid Server, the vault holds data that can be made available to AI systems under explicit, revocable consent.
Charlie sits on top of that vault as an intelligent layer between the individual and any LLM.
When a user makes a request, Charlie first identifies which data in the vault is relevant to that query and composes it to give the LLM the context it needs to respond well. Second, it strips out personally identifiable information before anything leaves the vault. Third — and this is the technical detail that tends to surprise people — it obfuscates the remaining data: close enough to the original that the LLM returns a useful, personalized response, but diffuse enough that the model never gets a precise bead on the individual.
"You can enjoy the experience of any LLM you want," John said, "without putting yourself in harm's way of them knowing you so intimately they can begin to use that to their advantage."
Critically, Charlie doesn't take your data. Inrupt can't see your data. The vault remains under the control of a trusted entity — a bank, a government, a telecommunications provider — and access to its contents flows only with your consent.
Charlie is coming. It will be available through the banks, telcos, retailers, and other high customer loyalty organizations who deploy Inrupt's Enterprise Wallet Infrastructure. To register your interest and be among the first to become a trusted partner, reach out below.

.jpg)
