Eventually, it’s bound to occur. AI systems will begin functioning as representatives, performing tasks on our behalf with a certain level of independence. I believe it’s crucial to consider the security aspects of this now, while it’s still an emerging concept.

In 2019, I became part of Inrupt, a firm that is bringing Tim Berners-Lee’s open protocol for decentralized data ownership to market. We are developing a digital wallet that aims to utilize AI in this manner. (We previously referred to it as an “active wallet.” Now we’ve dubbed it an “agentic wallet.”)

I discussed this briefly during my keynote at the RSA Conference earlier this week, focusing on AI and trust. Any effective AI assistant will necessitate a level of accessibility—and thus trust—that rivals what we currently expect from our email providers, social networks, or smartphones.

This Agentic Wallet serves as an illustration of an AI assistant. It will integrate personal data about you, transactional information you are involved in, and broader knowledge of the world. It will utilize this information to respond to queries, forecast outcomes, and ultimately act on your behalf. We have demonstrations of this functionality available right now. At least in its initial phases. Making it operational will demand an immense amount of confidence in the system. This necessitates integrity, which is why we are implementing safeguards from the outset.

Visa is also contemplating this concept. It recently introduced a protocol that employs AI to assist individuals in making purchasing choices.

I appreciate Visa’s method as it promotes an AI-agnostic standard. I have significant concerns regarding lock-in and monopolization within this sector, so any initiative that enables individuals to transition seamlessly between AI models is favorable. Furthermore, I like that Visa is collaborating with Inrupt to ensure the data remains decentralized. Here’s our announcement concerning their announcement:

This isn’t a fresh partnership—we’ve been collaborating for over two years. We’ve completed a successful proof of concept and are now establishing a sandbox within Visa so that merchants, financial organizations, and LLM providers can experiment with our Agentic Wallets alongside the full suite of Visa’s Intelligent Commerce APIs.

Moreover, we invite any other company wishing to delve into personal, consent-based Agentic Commerce to collaborate with us as well.

I joined Inrupt years ago because I believed that Solid could revolutionize personal data in the same way HTML transformed published content. I was drawn to the protocol’s open standard and its decentralization of data rather than its centralization. AI agents require decentralized information. The term “wallet” serves as an apt metaphor for personal data repositories. I’m hopeful this marks another advancement toward widespread adoption.