Opinion, Opinion You self-custody your bitcoin. Why not control your AI agent too? David Minarsch says we shouldn’t be trusting important tasks to rental agents. 

Who truly controls your AI assistant? That’s a question most people haven’t asked yet. Today, millions rely on digital assistants, from voice-controlled devices to smart bots embedded in tools like Google Workspace or ChatGPT. These systems help us write, organize, search, and even think. However, the vast majority of them are rented. We don’t own the intelligence we depend on. That means someone else gets to control it.

If your digital assistant disappears tomorrow, can you do anything about it? What if the company behind it changes the terms, restricts functionality, or monetizes your data in ways you didn’t expect? These are not theoretical concerns. They’re already happening, and they point to a future we should actively shape.

David Minarsch is a speaker at Consensus 2025 in Toronto May 14-16.

As these agents become embedded in everything from our finances to our workflows and homes, the stakes around ownership become much higher. Renting is probably fine for low-stakes tasks, like a language model that helps you write emails. However, when your AI acts for you, makes decisions with your money, or manages critical parts of your life, ownership isn’t optional. It’s essential.

What Today’s AI Business Model Implies for Users

AI as we know it is built on a rental economy. You pay for access, monthly subscriptions, or pay-per-use APIs, and in exchange, you get the “illusion” of control. However, behind the scenes, platform providers hold all the power. They choose what AI model to serve, what your AI can do, how it responds, and whether you get to keep using it.

Let’s take a common example: a business team using an AI-powered assistant to automate tasks or generate insights. That assistant might live inside a centralized SaaS tool. It might be powered by a closed model hosted on someone else’s server — and running on their GPUs. It might even be trained on your company’s own data — data you no longer fully own once uploaded.

Now, imagine that the provider begins prioritizing monetization, like Google Search does with its advertising-driven results. Just as search results are heavily influenced by paid placements and commercial interests, the same will likely happen with large language models (LLMs). The assistant you relied on changes, skewing responses to benefit the provider’s business model, and there’s nothing you can do. You never had true control to begin with.

This isn’t just a business risk; it’s a personal one, too. In Italy, ChatGPT was temporarily banned in 2023 due to privacy concerns. That left thousands without access overnight. In a world where people are building increasingly personal workflows around AI, this weakness is unacceptable.

On the issue of privacy, when you rent an AI, you often upload sensitive data, sometimes unknowingly. That data can be logged, used for retraining, or even monetized. Centralized AI is opaque by design, and with geopolitical tensions rising and regulations shifting fast, depending entirely on someone else’s infrastructure is a growing liability.

What It Means to Truly Own Your Agent

Unlike passive AI models, agents are dynamic systems that can take independent actions. Ownership means controlling an agent’s core logic, decision-making parameters, and data processing. Imagine an agent that can autonomously manage resources, track expenses, set budgets, and make financial decisions on your behalf.

This naturally leads us to explore advanced infrastructures like Web3 and neobanking systems, which offer programmable ways to manage digital assets. An owned agent can operate independently within clear, user-defined boundaries, transforming AI from a responsive tool to a proactive, personalized system that truly works for you.

With true ownership, you know exactly what model you’re using and can change the underlying model if needed. You can upgrade or customize your agent without waiting on a provider. You can pause it, duplicate it, or transfer it to another device. And, most importantly, you can use it without leaking data or relying on a single centralized gatekeeper.

At Olas, we’ve been building toward this future with Pearl, an AI agent app store realised as a desktop app that allows users to run autonomous AI agents with just one click while retaining full ownership. Today, Pearl contains a number of use cases targeting primarily Web3 users to abstract the complexity of crypto interactions, with an increasing focus on Web2 use cases. Agents in Pearls hold their own wallets, operate using open-source AI models, and act independently on the user’s behalf.

When you launch Pearl, it’s like entering an app store for agents. You can pick one to manage your DeFi portfolio. You can run another that handles research or content generation. These agents don’t need constant prompting; they’re autonomous and yours. Go from paying for the agent you rent to earning from the agent you own.

We designed Pearl for crypto-native users who already understand the importance of owning their keys. However, the idea of taking self-custody of not just your funds but also your AI scales far beyond DeFi. Imagine an agent that controls your home automation, complements your social interactions, or coordinates multiple tools at work. If those agents are rented, you don’t fully control them. If you don’t fully control them, you’re increasingly outsourcing core parts of your life.

This movement is not just about tools; it’s about agency. If we fail to shift toward open, user-owned AI, we risk re-centralizing power in the hands of a few dominant players. But if we succeed, we unlock a new kind of freedom, where intelligence is not rented but truly yours, with each human complemented by an “army” of software agents.

It’s not just idealism. It’s good security. Open-source AI is auditable and peer-reviewed. Closed models are black boxes. If a humanoid robot is living in your home one day, do you want the code running it to be proprietary and controlled by a foreign cloud provider? Or do you want to be able to know exactly what it’s doing?

We have a choice: We can keep renting, trusting, and hoping nothing breaks, or we can take ownership of our tools, data, decisions, and futures.

User-owned AI isn’t just the better option. It’s the only one that respects the intelligence of the person using it.

READ MORE: Olas’ Mech Marketplace Enables AI Agents to Hire Each Other for Help

 CoinDesk: Bitcoin, Ethereum, Crypto News and Price Data Read More