WWW Creator Tim Berners-Lee Unveils Privacy-First AI Agent

PARIS – More than three decades after inventing the World Wide Web, Tim Berners-Lee warned that artificial intelligence risks concentrating power in the hands of technology companies and stripping individuals of control over their personal information.

Speaking at the VivaTech conference in Paris alongside John Bruce, co-founder and CEO of Inrupt, Berners-Lee argued that the web’s original promise of empowerment is under threat as consumers increasingly rely on large language models and centralized platforms.

“When the web started, anybody could make their own website, and that was very empowering,” Berners-Lee said at the tech summit, which sponsored The AI Innovator‘s trip. “Now everybody’s on Facebook, people are using LLMs to ask questions where they could use a search engine just as well, so we’ve got disempowered people.”

The discussion centered on what the speakers described as a growing crisis of “individual sovereignty” in the AI era. While governments and businesses increasingly talk about national AI sovereignty and control over strategic technologies, Berners-Lee said individuals should have similar authority over their own information.

“An individual ought to have complete control of their data,” he said.

Bruce argued that the rise of generative AI has intensified privacy concerns because AI assistants are becoming repositories of highly personal information.

“With the advent of the LLMs, it’s becoming increasingly clear that they’re going to be your memory,” Bruce said. “The amount of data individuals are providing to the LLMs – you get a good return from the investment, you get good service – but the tradeoff is that they become your memory.”

He said users are often unaware of how much information AI systems accumulate about them over time.

“The degree of intimacy that the LLMs are beginning to build about you is way beyond where it ever was with search terms,” Bruce said.

Introducing privacy guard Charlie

To address those concerns, Berners-Lee and Bruce discussed a new AI privacy system called Charlie, developed by Inrupt, the company they co-founded. The system acts as an intermediary between users and AI models such as OpenAI’s ChatGPT, Anthropic’s Claude and other large language models.

The concept builds on Berners-Lee’s long-standing work on data ownership and Solid, an open-source framework designed to give users control over personal information through individual data stores.

Bruce explained that Charlie evaluates what information is needed to answer a query, removes personally identifiable information and sends only an approximation of relevant data to an AI model.

Before a prompt is submitted, he said, Charlie determines which data points are useful and packages them while stripping out identifying information. It can also “obfuscate” information by slightly altering details so users receive useful guidance without revealing exact personal data.

One example involves financial advice. Rather than sending exact bank balances, credit scores or account details to an AI model, Charlie can provide approximated figures that still allow the model to generate recommendations.

“It doesn’t send your actual balances. It doesn’t send your actual credit scores,” Bruce said. “It sends a little approximation.”

The approach reflects a broader concern that AI companies may retain vast amounts of personal information indefinitely. Asked whether large AI providers are storing user data despite privacy assurances, Bruce pointed to examples such as dynamic pricing and online tracking practices as evidence that data can be used in ways consumers neither expect nor understand.

The executives also argued that banks, insurers and other trusted institutions may have incentives to support privacy-preserving technologies. As AI agents increasingly mediate consumer interactions, traditional companies risk being disintermediated by large AI platforms that become the primary interface for financial, retail and other services.

“If I’m a bank,” Bruce said, “all the things that I normally grant to my customer – counsel and guidance on financial matters – all that gets swept away by ChatGPT.”

Berners-Lee framed the issue in philosophical terms. Rather than today’s “attention economy,” he said AI could help create what he called an “intention economy,” where individuals voluntarily share information to achieve specific goals while retaining ownership of the underlying data.

“I might want to share it later,” he said, describing travel preferences or personal plans. “But everything starts off private.”