OpenAI founder Sam Altman is the most recognizable name in the artificial intelligence (AI) space. Since the launch of ChatGPT in late 2022, AI has steadily crept into every corner of daily life, and Altman has emerged as the leading figure driving that shift.
This feature is a part of CoinDesk’s Most Influential 2025 list.
STORY CONTINUES BELOW
As AI has expanded and improved, it has become increasingly intertwined with crypto. A wave of decentralized applications and protocols now use AI to enhance or automate DeFi activity. At the same time, a growing group of builders believes the relationship goes both ways: AI may reshape crypto, but blockchains may also help fix some of AI’s emerging weak points, including compute, privacy, and identity.
The growing need for compute, driven in large part by the increasing use of generative AI, is one of the core issues that some in the crypto ecosystem have predicted as posing a problem in the imminent future. As our reliance on AI increases, more energy and compute is needed to keep systems running smoothly. In order to avoid a single point of failure, AI compute will need a massive, globally distributed network, which crypto could help coordinate.
“Where blockchain shines [in addressing compute] is effectively marketplaces and coordination, and so crypto definitely has a really powerful role here to tap into underutilized compute: how to get the best price, how to secure that compute, and provide confidentiality,” said Illia Polosukhin, the creator of NEAR Protocol. NEAR is a blockchain designed for fast, low-cost, developer-friendly applications, and recently the team behind it has expanded its efforts to build out AI tools, letting blockchain developers run AI products onchain. Polosukhin was one of the authors of a whitepaper that many see as the framework for modern-day LLMs, also known as the Transformer.
Today, essentially most AI development depends on a few large companies like Amazon, Google, and Microsoft/OpenAi, that control expensive, limited GPU resources, and there’s no easy way to coordinate or trust thousands of individual machines scattered around the world. Blockchains can come in and act as a neutral coordination and verification layer, recording which tasks were assigned, confirming whether they were completed correctly, and automatically paying the person who provided the compute. Because the records on blockchains are tamper-proof, users don’t need to trust a random machine owner; the blockchain proofs and transparent logs handle that.
In short, blockchain adds the trust, coordination, and incentive layers needed to turn millions of independent machines into a global network capable of powering AI.
There are a number of projects, known as decentralized AI networks, that have evolved from this market. One of the earliest ones is Bittensor, which provides a marketplace for computation.
The rise of decentralized AI networks is rooted in a growing frustration among developers, researchers, and crypto-native builders with how centralized and permissioned the AI ecosystem has become. Their concerns range from the concentration of compute and data inside a handful of corporations, to the lack of transparency in how models are trained, to fears that such centralized control could enable censorship, gatekeeping, or unilateral decision-making about which AI systems the world is allowed to use.
While Bittensor began as an AI project that used blockchain as a coordination tool, its founders say it has since expanded beyond just AI.
At its core, the Bittensor network creates an open marketplace for intelligence and computation: participants run models or supply hardware, and the network continuously evaluates the quality of their contributions. When a model produces something valuable, it earns the protocol’s native token, TAO. Over time, Bittensor has organized itself into specialized “subnets,” each focused on a different category of AI work. The result is an ecosystem that behaves less like a single system and more like a living environment, where many forms of intelligence evolve simultaneously. Good contributions rise, weak ones fall away, and anyone with skill or computation can participate without asking permission.
The push for decentralization of AI, according to Bittsensor co-founder Ala Shaabana, comes from what he describes as a structural imbalance in how modern AI is built and controlled. Today, virtually all meaningful AI power sits within a tiny handful of corporations.
“It’s like if three people in the world owned all the libraries, teachers and computers, and everyone else needed permission to use them,” he told CoinDesk in an interview. The concentration is so prevalent, Shaabana pointed out that OpenAI had just two board members empowered to make decisions about a technology the company itself equated to “the next best thing after a nuke” (though this was at the founding of the organization). To Shaabana, the notion that a small group of powerful individuals could unilaterally steer the development of something as consequential as AI is dangerous.
This is where crypto comes into play. Incentives are what make it possible to coordinate a global network of contributors who train models, provide data, and supply compute.
Though the average crypto investor has long considered privacy to be important, the issue of privacy has emerged as a growing concern among AI users this year. User data is often retained by and used to train the LLMs that power major AI platforms, leading to a host of concerns about how private, personal data could be used.
Polosukhin believes this is where “private AI,” or what he calls “user-owned AI,” becomes essential. The idea is that AI systems should operate on behalf of users or organizations within their own infrastructure, rather than sending sensitive data to centralized providers. This allows teams to train models for their specific needs while keeping information under their control and maintaining compliance with international privacy regulations like HIPAA and the European Union’s GDPR. Blockchains can provide tamper-proof logs and trust guarantees to support this framework.
Polosukhin argues that achieving user-owned AI requires rebuilding the AI stack itself — from compute to privacy to model training — in a way that puts control back into the hands of users and organizations.
“For that, you need a decentralized compute network, you need private AI, and you need model training,” Polosukhin said.
Beyond privacy, trust and identity also become more complex in the age of AI, which is another area where blockchain may again play a role.
Sam Altman’s controversial blockchain project, the World network, aims to address identity through proof-of-personhood (sometimes known as proof-of-humanity). The system gives users a World ID, a digital credential proving they are a unique human. Using its Orb device, the system scans a user’s iris to create a one-of-a-kind cryptographic code. According to World, the iris image is deleted, but the code remains so the system can ask if it is the same user as before: without revealing a name or any personal details, users can verify their identity online.
Tiago Sada, Head of Product at Tools for Humanity (TFH), the organization helping steer the World network, told CoinDesk that Altman identified identity and trust as core issues created by AI, and saw blockchain as a natural fit to address them.
“One of the things you lose in the age of AI is the ability to trust things online,” Sada said. “You don’t know who or what to trust. That’s where proof-of-humanity comes in. Whether you’re talking about tweets, a picture, someone sending money — blockchains can be a source of truth in a world where it’s really hard to know what truth is.”
Sada described World as a necessary safety layer, comparing it to inventing seatbelts alongside the invention of cars. He believes identity and privacy-protecting financial tools will become critical use cases at the intersection of AI and blockchain.
Beyond just verifying identity, Sada thinks that all of this will be essential for financial identity and use cases, to protect that data while not revealing anything about financial transactions, will be key to the intersection of AI and blockchain.
As speculation grows about whether we are in an AI bubble, none of the experts interviewed wanted to predict where things will land.
Of today’s AI projects, Sada of TFH estimated: “70% of it will go away — it was a fad. 30% of it is incredibly profound and will change the world. And that 30% is more than worth the hype of the others.”
Polosukhin, meanwhile, is concerned about the economic model that could emerge once AI becomes fully integrated into society. “As things become more efficient, capital returns get better, but labor loses access to capital,” he said. “We are going into something that doesn’t have an economic theory to function. There’s no clear model of how society works when only a small percentage of people oversee machines.” Crypto, he noted, offers a sandbox to experiment with new economic systems in ways traditional economies cannot. That idea is already being tested in the industry: Coinbase recently launched a universal basic income pilot, using blockchain rails to distribute recurring payments and explore how crypto-based economic mechanisms could support people.
Reflecting on the widespread use of modern-day LLMs that Polosukhin helped design, he finds it “exciting” seeing the subject he worked on come to life. He added, “It’s great to see this now fully working. Obviously there’s still a lot to improve, but there’s been a massive shift away from what was considered machine learning 15 years ago to what we have now.”