Vision of decentralised control has drawn other major players to the effort
Sooner or later, artificial intelligence is going to take on a new role: that of an autonomous agent, acting on behalf of individuals with little or no direct input. That future may seem distant now, but it’s approaching fast—and some technologists are already building the safeguards they believe will be crucial to earning public trust in such systems.
One of them is Dr. Pankaj Sharma, a Bengaluru-based cybersecurity strategist and AI governance expert, who believes that the conversation around digital autonomy and trust needs to begin now. Sharma is working with Inrupt, a company co-founded by web inventor Sir Tim Berners-Lee, which is developing technology based on the open Solid protocol—an effort to return control of personal data to individuals. Now, Inrupt is developing what it calls an “agentic wallet”: a digital assistant designed to act on a user’s behalf using AI, while respecting their privacy and consent.
“Any useful AI assistant will need access—and therefore trust—that matches what we give our email provider, smartphone or cloud account,” Sharma said during a recent conference on AI and trust in New Delhi. “We cannot afford to build this layer without first embedding security and decentralisation into its core.”
The concept is ambitious. The agentic wallet will combine personal information, transactional data and general knowledge about the world to answer questions, make predictions, and eventually act. Think of it as a personal data vault that doesn’t just store information, but uses it intelligently. Demos of the technology are already running in early-stage environments.
But for such a system to succeed, Sharma says, it must operate with an unusually high level of trust and transparency. “We are designing it in a way that ensures integrity from day one. There is no other path forward.”
That vision of decentralised control has drawn other major players to the effort. Visa, the global payments company, recently announced a protocol that uses AI to help consumers make purchasing decisions. Importantly, Visa is working alongside Inrupt to ensure that user data remains decentralised—an uncommon move in the typically centralised payments ecosystem.
“I see Visa’s approach as a positive signal,” Sharma said. “It’s AI-agnostic, and it promotes competition by making it easier for people to switch between AI models. That helps reduce the risk of monopolies, which is a serious concern.”
The collaboration between Inrupt and Visa isn’t new. The two have been working together for over two years. Following a successful proof-of-concept, they are now setting up a sandbox environment inside Visa to allow merchants, financial institutions and LLM (large language model) providers to test agentic wallets alongside Visa’s Intelligent Commerce APIs.
The call is open for others as well. “We welcome companies interested in personal, consent-based agentic commerce to join this effort,” said Sharma.
At its heart, the agentic wallet is an extension of a long-standing vision: that individuals, not corporations, should control their data. “The Solid protocol is to personal data what HTML was to online publishing,” Sharma said. “AI agents will only thrive if they are built on decentralised data infrastructures. The term ‘wallet’ is more than metaphor—it’s a functional representation of data autonomy. This could be a defining step toward that future.”
As autonomous AI systems edge closer to everyday use, the focus will shift from what these systems can do to who controls how they do it. For Sharma and others working in the space, the answer lies in openness, user control, and trust from the very beginning.
