Panel Report from Futurist, Toronto
The battle for control of artificial intelligence is no longer theoretical. It’s real, escalating, and already shaping the future of human agency. In the age of agentic systems and GPU monopolies, decentralization is not a luxury but a necessity. What happens when a few actors hold the keys to compute, data, and model access? This panel explored how infrastructure, privacy, and access must evolve to ensure AI remains a tool for collective progress, not private control.
Panelists:
Moderator: Michael Casey, Chairman, Decentralized AI Society
Lisa Loud, Executive Director, Secret Network
Mark Rydon, Co-Founder & CSO, Aethir
Tony Evans, Co-Founder & CSO, CETI
Casey opened by framing the moment as existential:
“We are in the biggest fight of our lives to decentralize AI and secure humanity’s future.”
He pointed to compute as the base layer, not just a resource but a strategic battleground, and introduced the panel as a coalition of builders addressing this challenge from different fronts: confidential computing, decentralized infrastructure, and access-based economics.
Loud explained Secret Network’s use of Trusted Execution Environments (TEEs) to embed encryption directly into the chip layer. “We’re building the foundational layer so that everyone else can build AI that is confidential and secure,” she said. Secret’s approach enables encrypted data to be stored on-chain, with view permissions controlled by the user, a critical privacy innovation for agentic AI.
Evans, whose background spans traditional finance and crypto, described CETI’s mission as one of accessibility: “We literally said, if compute is controlled by a handful of people, people won’t have access. So we invested in the hardware and made it available to anyone building decentralized AI.” For him, decentralization is about market freedom, the right for entrepreneurs to build without centralized gatekeepers.
Rydon emphasized GPU scarcity and the risks of monopolization. Aethir aggregates enterprise-grade GPUs globally, creating a cloud-like layer without owning the hardware itself. He recalled Elon Musk’s recent goal to build a $25 billion data center with a million GPUs: “Nvidia doesn’t just hear that and go produce an extra million. Those GPUs are now unavailable to the rest of us.”
Evans agreed: “The only reason we got GPUs was not because we asked nicely. It was because we knew somebody. The power Elon would have with a million GPUs is closer to a medium-sized nation.”
As the discussion evolved, Casey pushed on the concept of democratization: “How do we get to a place where I can have my own AI agent , even if the compute lives in someone else’s data center?”
Loud’s response was quick: “We’re in a fight. People are gathering power in one place. But we should each be controlling our own data, our own agents.”
That’s only possible, she added, with confidential systems that allow private data to be shared selectively and securely, not dumped on a transparent blockchain.
The panel pivoted to economic incentives. Could decentralized AI infrastructures really compete with centralized powerhouses like OpenAI or AWS? Evans was unequivocal: “Yes. If we didn’t believe that, we should just pack up now and sod off.”
Evans advocated for a “threesome” model, integrating physical, digital, and AI elements, and warned that our current laissez-faire attitude to data was a mistake: “Information isn’t power. People who can apply it are.”
Rydon added nuance. In pure performance terms, centralized systems win, for now. “You need adjacent compute, expensive proprietary data, and speed,” he said. But there’s hope: “At the application layer, latency matters. That’s where distributed compute can shine. Not every problem needs a supercluster.”
The conversation turned speculative: what happens when the global power grid caps out? Rydon sees that as a potential decentralizing force. “We’ll need to plug into latent compute, from gaming PCs, from mobile devices, or shift toward location-specific renewable sources.”
Casey steered the panel toward opportunity: could privacy-enhanced AI agents unlock the next wave of innovation? Loud was optimistic. “We’ve hit the limit of what we can do with public data. But if we protect private data, we can unlock breakthroughs in medicine, climate, even curing cancer.”
In closing, Evans brought it back to the human element. “The beauty of humanity is imperfection. Centralized AI pushes toward a single monolithic output. But creativity comes from variability, from different minds working in different ways. We lose that, we lose everything.”
Rydon agreed, noting that blockchain-based AI has the potential to allow model transparency and data accountability, including exclusion, not just inclusion.
“Let’s decide what models we want to build,” he said.
The fight to decentralize AI is far from over, but this panel showed that the infrastructure is forming, the will is there, and the case for agentic, private, and equitable AI has never been clearer.
