Charlie Is an AI Assistant Built to Serve You, Not Our AI Overlords

It's entirely possible that Tim Berners-Lee is the most trustworthy man in tech.

Amid the wider landscape of money-motivated tech entrepreneurs, the inventor of the web made his creation free and open, so that it would be accessible to all. And even today, in the age of billionaire tech bros, he stands by the decision not to sell out to cash in on his work. "It's much more fun to have the web," he told the audience at SXSW London on Wednesday.

The 70-year-old Berners-Lee has a legacy of building tech designed for the good of humanity, and he's still at it today. Together with John Bruce, the CEO of Berners-Lee's company, Inrupt, he introduced the SXSW audience to his AI agent, Charlie.

Charlie isn't actually a new idea -- it was originally proposed by Berners-Lee around a decade ago as an alternative to the Apple-owned chatbot Siri and the Amazon-owned Alexa. But unlike those assistants, which ultimately work for their Big Tech overlords, Charlie is an AI-powered assistant designed to serve you and only you.

Given that an AI agent is a system that can make decisions and take actions on your behalf, trust is absolutely imperative here. It would be so easy while interacting with popular LLMs or chatbots -- whether that's ChatGPT, Claude, Gemini or some other agentic assistant -- to leave messy trails of your personal and private information all over the internet if they have unfettered access to your data.

Charlie is different. There are several steps to how it works, but to lay it out simply, imagine that all your data is stored in a vault owned and controlled by you. Charlie is effectively the gatekeeper to that vault and will interact on your behalf with any LLMs that request access to your data. Before it hands anything over, it will first ask for your permission.

If you grant permission, it will perform obfuscation -- changing little details -- before it hands your data over to the service requesting it. This way, the LLM will have enough information to give you an accurate answer, but not to get a read on you and who you are.

Most people are likely to encounter Charlie first through their bank, Bruce told me in a follow-up interview at SXSW. But he "absolutely" thinks that at some point it will become an app you can have on your phone, he added.

The timing of the technology feels critical -- potentially already overdue. People are already liberally uploading financial information to LLMs, Bruce said. The intimacy with which they then know you is frightening, he added -- and they never forget.

At a time when our trust in the big tech companies of the world -- motivated as they are by profit -- feels fragile, knowing that Charlie was created by Berners-Lee feels reassuring. His involvement, though, isn't the only, or even the main, reason to trust Charlie, according to Bruce.

The trust in the technology will be built as it's used by the institutions we already trust with our most precious assets, including our banks, but in some cases, government bodies or other entities.

Right now, Inrupt is starting to work with major partners to get Charlie out into the world so it can help us protect sensitive data, such as our financial and health information. With the right momentum, Charlie might eventually be a reliable presence on our phones, giving us confidence that we're in charge of our own digital destinies.

After many years of talking about Charlie's values and hammering out how it should operate, Berners-Lee said he's "right chuffed" (that's British for very pleased) to have the agent out in the world.

As the biggest AI companies in the world prepare for their IPOs, it's some comfort to know that at least one of the top minds in tech is still thinking about people rather than big numbers.