Panel report from Futurist, Toronto
In a world where AI can act on your behalf, the stakes around privacy are no longer theoretical. They are architectural. The panel “Privacy and the Agentic Future,” moderated by Cointelegraph’s Sam Bourgi at the Futurist Conference in Toronto, brought together leading minds in confidential computing, agentic AI, and digital security. Together they examined how decentralised systems can support increasingly autonomous agents, without compromising user sovereignty.
Panelists
Sam Bourgi (Moderator), Journalist, Cointelegraph
Lisa Loud, CEO, Secret Network Foundation
Todd Ruoff, Founder, Autonomys Network
Shaw Walters, CTO, Eliza Labs
Yevheniia Broshevan, Co-Founder and CBDO, Hacken
The first question posed by Bourgi focused on what’s at stake as agents evolve from tools into decision-makers. “We’re entering a phase where AI can act without asking,” said Shaw Walters. “That means it needs boundaries.”
Eliza Labs is building agent infrastructure that operates across protocols, giving users control of their data while letting AI act autonomously within pre-set limits.
“Today’s agents are like toddlers with a wallet. They need safety rails,” Walters said. The analogy set the tone for the entire session. With AI tools increasingly embedded in daily decision-making, from asset allocation to identity verification, the lack of native constraints poses more than just technical risk. It becomes an existential one.
Lisa Loud argued that privacy is essential to that process. “We’re building systems that think and act on behalf of users. If they’re going to do that, privacy must be embedded, not bolted on,” she said. Secret Network’s architecture encrypts inputs, logic, and outputs. That allows agents to act intelligently without leaking sensitive information, a requirement that grows more urgent as these systems scale.
Todd Rouff added that agentic autonomy is not just about offloading tasks. “It’s about redesigning infrastructure for sovereignty,” he said. Autonomys Network enables users to deploy their own intelligent agents across platforms, agents that negotiate, stake, and execute without sacrificing control. “If you don’t design for sovereignty now, you’ll rebuild the same power structures later.”
Bourgi then shifted the discussion to risk and regulation. Yevheniia Broshevan of Hacken warned of reputational fallout if privacy isn’t handled well. “We audit these systems every day. We know how fragile they can be,” she said. “Privacy without security is meaningless.”
She pointed out that the public perception of privacy tech remains polarised. “After Tornado Cash, many policymakers are wary of any system that conceals on-chain behaviour,” she said. “But privacy and compliance are not mutually exclusive. What we need is verifiability and that means building with transparency and auditability from day one.”
All panelists agreed: agentic systems need infrastructure that supports privacy, auditability, and user control. Walters described Eliza’s use of economic constraints within agents, allowing them to execute decisions within a predefined sandbox. Loud explained how Secret’s SNIP-based encrypted tokens preserve transaction confidentiality while remaining traceable for compliance purposes. Broshevan advocated for audit trails, penetration testing, and red-team simulations prior to agent deployment.
“It’s not about hiding,” Loud clarified. “It’s about choosing what you reveal. That’s the foundation of trust.”
Rouff returned to the issue of adoption. “If users feel watched or unsafe, they won’t delegate authority to machines. They’ll stick with what they know,” he said. That, in his view, is the biggest barrier to innovation. “People need to know the tools are working for them, not watching them.”
The discussion also touched on the rapidly evolving regulatory landscape. Broshevan emphasised the role of ethical design in influencing policy. “If we wait for regulation, it may be too late. But if we lead with secure, user-first products, we can shape the conversation.”
There was also a sense of urgency. As AI systems grow more autonomous and financialised, the window to embed the right structures is closing. “The next twelve months will define how agentic systems are perceived,” said Walters. “If we get it wrong now, we may not get another chance.”
The panel ended with a real-world example. A global consumer brand, facing surveillance risks when managing cross-border campaigns, used Secret Network’s infrastructure to deploy regional marketing agents. These agents operated autonomously, negotiating spend, targeting segments, and executing creative strategy, without exposing campaign metadata or financial flows. “That’s the model,” said Loud. “Autonomy without exposure.”
As agents gain capabilities, the guardrails must be native. Privacy is no longer a feature to add. It is the context in which trust can exist. And in the agentic future, that context may be all that stands between innovation and abuse.
AI may be learning how to act. But the question is whether we’ve taught it how to respect.
