Web3 Meets AI Agents: What Will Matter In The Next 3 Years?

In Brief

A standout panel at the Hack Season Conference explored how blockchain could reshape AI’s future—through decentralized compute, storage, identity, and data attribution—offering real solutions to centralization, trust, and scalability challenges.

July is officially behind us. So, it’s a great moment to look back at one of the most forward-thinking panels of the month — a standout from the Hack Season Conference in Cannes. The session dove deep into the collision of blockchain and AI agents, bringing together some of the most technically ambitious founders at this intersection. The discussion didn’t hold back — tackling where things stand, what’s broken, and where this space could realistically be headed.

The panel featured:

• Tomer Sharoni, CEO and Co-Founder of Addressable;
• Tom Trowbridge, Co-Founder of Fluence;
• Mathieu Baudet, Founder & CEO of Linera;
• Clara Tsao, Founding Officer of Filecoin Foundation;
• Kamesh, Core Contributor of OpenLedger.

Moderated by Tomer, the session cut through hype to examine how blockchain might truly reshape the future of AI—and what it will take to get there.

Decentralized Compute: The Foundation for Scalable AI

Tom Trowbridge kicked off with a simple but compelling premise: the AI revolution isn’t just about smart models—it’s about access to infrastructure. Fluence is building a decentralized compute network that starts with CPUs and is expanding to GPUs. Why? Because right now, large centralized companies are locking up most of the world’s compute power, creating enormous barriers to entry for smaller players.

He argued that AI is rapidly becoming a central source of truth in our digital lives—much like search engines or news aggregators once did. And as that power concentrates in a few hands, it becomes vulnerable to bias, censorship, and influence. Blockchain, according to Trowbridge, offers a way out of this trap by decentralizing not just the training and inference layers, but the very idea of AI truth itself.

In his view, the most trusted AIs of the future won’t be built by the richest companies—but by networks without shareholders, without board pressure, and without centralized control.

Data Trust and the Storage Layer

Clara Tsao from the Filecoin Foundation picked up this thread with a look at the data supply chain that feeds AI. If AI is going to generate most of the world’s digital content by 2026—as some estimates suggest—then how that data is stored, accessed, and verified becomes a foundational concern.

Filecoin’s decentralized storage layer is built for precisely this world. Clara stressed the importance of traceability, particularly as users begin offloading highly personal or sensitive information to AI agents. Whether it’s relationship advice, therapy notes, or corporate secrets, users often don’t know where their data is being saved or who can access it.

Filecoin aims to offer not just decentralization, but transparency—giving users confidence that their content is stored verifiably and immutably. She pointed to potential use cases ranging from scientific reproducibility to journalism and content attribution—sectors that are particularly vulnerable to deepfakes and AI-generated misinformation.

AI Needs Web3—but It Doesn’t Know It Yet

Mathieu Baudet of Lineara brought a candid and slightly contrarian perspective. He argued that despite the philosophical alignment, most AI companies don’t yet care about what Web3 can offer. They’re focused on scale, speed, and performance—not decentralization. But he believes that will change once security and integrity issues become too big to ignore.

According to Baudet, today’s blockchains weren’t designed for machine-to-machine interaction. AI agents that want to interact with blockchains securely often have to rely on centralized RPC providers—creating attack surfaces that undermine the whole premise of decentralization.

Lineara’s solution? Sparse clients that give users (and agents) the ability to locally store and verify relevant parts of the blockchain without the need to trust third-party intermediaries. This architecture allows for highly secure, low-latency interactions—exactly what AI agents will need when they start operating autonomously in financial markets, supply chains, or digital identity systems.

Who Controls the Data?

Kamesh from OpenLedger emphasized the often-overlooked problem of attribution. Right now, there’s no clear way to track how specific data sets contribute to an AI model’s behavior. That creates a massive disconnect: content creators fuel the training process, but they’re never compensated—and often, not even acknowledged.

OpenLedger’s vision is to make data provenance visible and enforceable. If you can trace how a model’s outputs were shaped by specific data, then you can start to build systems where contributors get paid—automatically and verifiably. This isn’t just a technical problem, Kamesh noted—it’s a legal and ethical one, and it’s already creating friction between AI companies and rights holders.

He believes blockchain is uniquely suited to solve it, especially as regulators and creators begin demanding accountability in how data is collected and monetized.

Agents as Consumers: Why the UX Doesn’t Matter Anymore

Returning to the role of AI agents, Tomer of Addressable posed a provocative idea: AI agents will become the most efficient consumers in the world. They don’t care about the brand, they don’t need sales reps, and they’re indifferent to poor UX—as long as the performance and pricing is right.

This has major implications for Web3 adoption. Many promising crypto projects struggle to get traction because they’re too technical or too clunky for mainstream users. But AI agents don’t need slick onboarding—they just need APIs. In this way, Tomer argued, AI agents might actually accelerate adoption of decentralized technologies by acting as super-users: autonomous, tireless, and laser-focused on efficiency.

He also warned that identity will become a critical issue as agents multiply. With countless non-human entities operating online, verifying human provenance—proving that someone on the other end is actually a person—will become a pressing challenge. Here too, blockchain might offer the most elegant solution.

What Will Blockchain Mean for AI in Three Years?

The panel closed with a rapid-fire look at the future. Each participant offered a vision of where blockchain’s intersection with AI will become indispensable:

Kamesh pointed to legal and enterprise use cases—especially agent traceability and model transparency—as likely early adopters. Clara emphasized decentralized storage as the essential base layer for a trustworthy AI future, with attribution, reproducibility, and journalistic integrity as critical use cases. Mathieu predicted the rise of embedded agents with direct, secure blockchain access—shaping everything from DeFi to smart infrastructure. Tomer focused on identity, suggesting that blockchain could be the key to distinguishing humans from agents as the digital world gets increasingly crowded with non-human entities.

One thing they all agreed on: the AI landscape is moving fast—and the problems of today will look quaint in just a few years.

To catch every insight from the panel, watch the full video here: AI & AI Agents: How to Win in the New Era