Rune Bentien
Rune Bentien

For years, the crypto industry waited for permission. It waited for regulators to say yes. It waited for banks to feel safe. It waited for the grown-ups to arrive.

The grown-ups are here. Stablecoins settle real payments. Banks are building digital asset platforms. Institutional capital is in the room. The technology has moved beyond proof of concept and into production. That fight is over.

But the shift was slow. It took more than a decade of consolidation, regulatory clarity, and institutional caution. The tourists left. The speculators moved on. What remained was infrastructure, and the people willing to build it. The real change has already happened. Now the machines are next.

AI agents are beginning to act on our behalf. They transact, negotiate, and execute tasks without a human approving every step. EY found that 48% of technology leaders are already deploying or adopting agentic AI. The law has not caught up. When an agent causes harm, no case law clearly says who is responsible. The model provider? The developer? The company that deployed it? The person who prompted it? Existing law simply does not answer this question.

Insurers do not solve this. They cover the company. They do not cover the agent. The Pearl Cohen Law Firm notes that the Ninth Circuit reasoned that the AI agent is a tool, not a person capable of accessing a computer, and expressly disclaimed establishing a new legal regime governing agentic AI. A wrong action still needs a human or corporate owner to carry the liability. That leaves a gap.

This is the moment to decide what kind of system we build. The industry can repeat its oldest mistake, or it can learn from it.

The old mistake was treating regulation as the main constraint. Waiting for clarity. Asking for permission. Playing it safe. That posture produced compliance, but it also produced inertia. The companies that built the internet of value did not wait for a rulebook. They built, tested, and pushed. Some broke things. But they moved.

The same temptation is here now. We can wait for regulators to define agent liability. We can let the largest platform companies write the rules through their terms of service. We can accept a future where a handful of firms control the models, the data, and the rails.

Or we can build differently.

Open source is not a slogan. It is an insurance policy. If a few companies control the models that agents run on, they control the terms of every transaction those agents execute. Open, locally controllable AI infrastructure keeps that power distributed. It lets businesses run agents on their own terms, with their own data, under their own oversight.

Liability must stay human. An agent can act. A person or a company must answer for what it does. Sandboxes can test agent-to-agent payments and delegated spending on blockchain rails without pretending regulation alone will solve the problem. Experimentation can help close the gap, but only if it is designed to produce accountability, not just speed.

And we should stop thinking small. Current use of AI is incremental. Drafting emails. Summarizing meetings. Automating expense reports. These are fine. They are not the point. These tools can attack problems that have resisted human effort for decades. Climate. Poverty. Financial access. If the industry stays focused on saving minutes, it will miss the chance to save years.

Preserve what crypto got right. Permissionless access. Open participation. Decentralization where it matters. Institutions entering the space should not mean the space becomes another closed, intermediary-controlled system. The principles were not naive. They were the reason the technology worked.

If we get this right, AI becomes an empowerment layer. People use agents to remove administrative burden, build products without traditional technical skills, and spend more time on human connection, creativity, and high-value judgment. Open systems keep a small group of giants from controlling data, models, and digital finance. Agents expand economic participation globally. Problem-solving compounds across climate, poverty, and access.

If we get it wrong, power concentrates further. Governments and incumbents shape the rules. Their incentives are profit and control, not openness or privacy. People feed personal data into systems they cannot inspect. Agents make consequential financial decisions without clear liability. Centralized companies capture the value from tools that should be open. AI deepens surveillance instead of empowering people, and open global finance becomes another closed platform.

That future is not inevitable. It is a choice.

One more thing. We groupthink. We chase hype. We fear missing out. That is human behavior, and it is why so much capital piles into the same ideas. Agents do not share that trait. They can spread across problems, test more paths, and explore ideas humans overlook. That is the real upside. Not faster emails. Tenfold or hundredfold problem solving.

The grown-ups have arrived. The machines are next. The question is not whether they will act on our behalf. It is who will answer when they get it wrong. If the answer is no one, we have built the wrong system.

About the Author

Rune Bentien is an investor and technology executive focused on blockchain infrastructure and digital assets. He is the founding partner of Lyrik Ventures. He works with founders building open, permissionless financial systems and writes on the intersection of regulation, artificial intelligence, and market structure. He is based in Europe.