Why Critical Infrastructure Needs Decentralized AI

Quick Take

Episode 5 of The White Papers, hosted by The Block's Gareth Jenkinson, explores the promise and challenges of decentralized AI with 0G Labs CEO Michael Heinrich.

Key topics

• 0G Labs combines blockchain, decentralized compute and AI models to provide verifiable, privacy-focused AI infrastructure.
• CEO Michael Heinrich believes rising costs at OpenAI and Anthropic could drive users toward open-source alternatives.
• Early use cases for decentralized AI include autonomous trading agents and tokens that generate compute credits.
• Autonomous AI systems present new societal risks; 0G Labs plans to develop new safety mechanisms and expand into robotics.

Timestamps
00:00 - Introduction
02:54 - 0G's tech stack and use cases
06:34 - OpenAI and Anthropic versus 0G
10:18 - The need for decentralized AI 
14:38 - AI in 2027 and beyond 
17:42 - Closing thoughts

Summary

With artificial intelligence increasingly integrated into critical financial, industrial and even military systems, questions of safety and transparency are at the forefront of the conversation. The "black box" systems of closed labs like Anthropic and OpenAI currently lead the industry, but some decentralized competitors are seeking to provide more transparent and secure alternatives using blockchain technology.

In episode 6 of The White Papers, recorded at Token2049 in Singapore, we spoke with the leader of one of those competitors: 0G Labs co-founder and CEO Michael Heinrich. In conversation with The Block's Gareth Jenkinson, he shared his thoughts on how decentralized AI could address some of the industry’s most pressing challenges.

Heinrich entered the blockchain and AI spaces after a long career in technology and management, spanning Microsoft, Bain & Company, Bridgewater Associates and other major firms. He and co-founder Thomas Yao started 0G in 2023 after witnessing the rise of ChatGPT. They predicted that these systems were soon "going to run more societal level use cases like running an airport, running logistics systems. We said we need an infrastructure that actually makes that possible."

Three years later, the company has developed a vertically integrated, blockchain-based tech stack for such solutions. This combines its own EVM L1 blockchain with decentralized compute and storage. The 0G network uses trusted execution environments to protect AI inference while recording verification data on the public blockchain.

Together, this allows developers to build AI-powered solutions that leverage blockchain's verifiable nature, while also keeping sensitive user data private. These dual benefits, Heinrich claimed, are essential for running safe, accountable AI agents.

He drew a comparison with the leading centralized providers such as Anthropic and OpenAI, whose systems are opaque: "You have no way of verifying what they do with your data, and so it's just a black box, effectively."

For this reason, early demand for 0G’s services is coming from institutional customers, he said, in industries such as healthcare where data protection is a paramount concern. Fears of data leaks, intellectual property theft and other security breaches are making these industries skeptical of adopting conventional AI solutions.

Heinrich added that another tailwind for decentralized AI could be the long-term economics of centralized AI. With some of the top providers aiming to go public in the near future, the pressure for them to stay profitable will only grow. This, he said, will lead to slashed subsidies and more restrictive subscription limits, ultimately driving users to open-source models.

"The price to value may not make sense anymore for many large corporations or other users. [...] They may start to say: the performance trade-off is very little on these open source models. I'm just going to be using these going forward," he said.

This is the future demand that 0G Labs aims to capitalize on. As for what activity is already taking place on the network, Heinrich highlighted some prominent applications covering AI-powered trading, agent creation, coding and DeFi. Ecosystem platform NeoSoul already hosts a reported 4 million agents on the network.

When pressed on what the future holds for 0G and the AI industry in general, Heinrich identified security as the most immediate concern. One underappreciated threat, he said, is so-called "recursive AI," wherein models train subsequent models themselves. This could allow harmful agents to proliferate unchecked, potentially threatening critical infrastructure.

Altogether, Heinrich views this as a validation of the mission 0G set out with in 2023: to provide stronger safeguards, transparency and accountability for the future of AI.

About the guest

Michael Heinrich is the co-founder and CEO of decentralized AI infrastructure company 0G Labs. Previously, he worked at Microsoft, SAP, Bain & Company and Bridgewater Associates before founding Garten, a Y Combinator-backed startup he later sold. A Stanford graduate, Heinrich entered the blockchain industry as an investor before co-founding 0G Labs.