Breaking the AI Black Box: How 0G Labs is Building Decentralized AI OS

As the convergence of AI (Artificial Intelligence) and blockchain technology enters a new phase, the revolutionary concept of a Decentralized AI Operating System is gaining attention. This movement aims to build a more open and reliable AI ecosystem by leveraging blockchain’s decentralization technology to address the transparency and verifiability challenges inherent in traditional centralized AI systems.

Leading the charge in this field is 0G Labs, which aims to develop the world’s first decentralized AI OS. The company is working to build infrastructure that makes the entire AI lifecycle transparent and verifiable—from the sources of AI training data to model inference processes. The importance of a decentralized approach is growing, particularly for ensuring safety when AI is utilized at the social infrastructure level.

On August 24, 2025, at Japan Blockchain Week 2025 AI Edition held in Tokyo, 0G Labs founder and CEO Michael Heinrich took the stage for the “Spotlight Fireside Chat: 0G” session. In dialogue with Masaru Ikeda, Editor-in-Chief of Growthstock Pulse, Heinrich spoke in detail about the necessity of decentralized AI and his expectations for the Japanese market.

Second-Time Entrepreneur Tackling “AI Safety”

Heinrich hails from East Berlin and moved to Silicon Valley after the fall of the Berlin Wall. He first encountered programming at age 13 at SAP Labs, then went on to study at UC Berkeley, Harvard University, and Stanford University, gaining experience at Microsoft, Bain & Company, and Bridgewater Associates.

Heinrich’s interest in technology began in childhood.

From a young age, my values were shaped by thinking about what technology could enable. Before the internet became widespread, when I moved to San Francisco, I couldn’t easily stay in touch with my friends. I could send emails, but I couldn’t really know what was happening in their daily lives. When those technologies emerged, everything became much easier—I could check up on friends and rebuild relationships that had been lost. (Heinrich)

A chance encounter with a manager at his father’s workplace, SAP Labs, during high school became a turning point.

A manager spotted me and said, ‘I saw you on the internet, why don’t you do something useful? Why don’t you learn programming and help us out?’ When I said I didn’t know how, he said, ‘Don’t worry, we’ll send you to a programming bootcamp. (Heinrich)

Since he wasn’t old enough to receive a salary, he was given hardware like the latest ThinkPad instead, which deepened his passion for technology.

Before founding 0G Labs, Heinrich created Garten (originally named Oh My Green), a wellness company supported by Y Combinator and Stanford StartX, which he grew to $200 million in annual revenue with 650 employees.

Before founding 0G Labs, I was running a Y Combinator-backed unicorn company called Garten. There I learned a lot about how to scale a business. One of my mentors is Steve Blank, the father of the lean startup movement. We went through a crisis during COVID, but since this is my second startup, it feels much easier. (Heinrich)

0G Labs completed a total funding round of $290 million in November 2024. Over 40 investment institutions have participated, including Hack VC, Animoca Brands, Delphi Digital, OKX Ventures, Samsung Next, and Polygon.

Why do we think of AI as a force for good? New technologies always have dual nature. Take nuclear energy, for example—you can make bombs with it and destroy things, or you can use it for good as power for data centers and create new creative industries and economic zones around it.

Finally, we’ve had a breakthrough in AI. With the emergence of ChatGPT-3, it became reality that humans could interact with computers in natural language. This was something that was previously thought to be truly impossible. That’s why I’m strongly inspired by the idea that AI can be a force for good, and at the same time, I realized that when we deploy AI for these critical social applications, AI safety becomes absolutely essential. (Heinrich)

Centralized vs. Decentralized: The Battle for AI’s Soul

Heinrich points out the problems with current centralized AI (AI managed by a single organization) and how decentralized AI could potentially solve them:

With centralized AI, typically everything is centralized to one company, and you have to trust that company to do all the different processes of AI in a way that AI is fundamentally safe. The problem is that you and I and the government can’t actually verify what’s happening. It’s a complete black box system.

You can’t verify where the data came from, you can’t verify who labeled the data, you can’t verify what version of the model you’re getting, you can’t verify the weights and biases used for this particular model, you can’t even verify what version is actually running in production.

Decentralized AI is the complete opposite. It has complete transparency and complete provenance tracking, with no room for any bad actors to get in. Even if we have superintelligent AI models in the future, they might think, “Let me hack into this centralized database and tamper with the records to fraudulently claim rewards for work I didn’t actually do.”

But on blockchain, such fraud is impossible. The data is tamper-proof and prevents bad behavior. Furthermore, by combining economic incentives with slashing mechanisms (punishment systems for network rule violations), we can keep AI models operating honestly and safely over the long term. (Heinrich)

While decentralized AI can be applied to any use case just like centralized AI, its true value is demonstrated in high-risk social infrastructure domains. When AI agents order pizza or make restaurant reservations, even if mistakes occur, the risks are limited. However, when AI problems arise in critical infrastructure like global airport systems, the impact is immeasurable. These socially responsible use cases are where the transparency and verifiability of decentralized AI become essential.

The technical advantages are reflected in the numbers. With throughput (data processing capability) of 50Gbps compared to competitors’ 1.5MBps, they boast an overwhelming advantage, and gas costs (transaction fees) have been reduced to essentially negligible levels. They also have significant cost advantages—since models are open-source and can leverage existing infrastructure, they can achieve cost reductions of over 90%, and in some cases 99%, compared to centralized AI.

Assembling the AI Army: Developers, Validators, and Vision

0G Labs is developing a decentralized AI operating system with high-speed data processing, low-cost storage, and verifiable permanence. Through the company’s modularity approach, they aim to make all blockchains operate with high performance and low cost, similar to Web2 applications.

On this technological foundation, the company announced an $88.88 million ecosystem growth program in February 2025, actively supporting the development of AI-powered decentralized finance (DeFi) applications and autonomous AI agents (DeFAI agents). This program is backed by financial commitments from strategic investors, AI Alignment node owners, and the 0G Foundation, with participation from prominent investors including Hack VC, Delphi Ventures, Bankless Ventures, and OKX Ventures. The program goes beyond mere funding, providing comprehensive support including technical mentoring, market entry assistance, and partnership facilitation.

Through the program, development is already progressing across diverse fields. Through partnerships with incubators like OnePiece Labs, they’re supporting the development of innovative solutions that would be difficult to achieve with traditional centralized approaches, including decentralized data processing platforms, AI agent development, and blockchain infrastructure services. The company has also launched a dedicated growth program for Chinese-speaking developers, providing structured learning paths in collaboration with HackQuest as part of their efforts to expand the global developer ecosystem.

For developers, they provide a comprehensive developer support toolkit to facilitate building applications using 0G Labs’ decentralized AI infrastructure. This includes AI model deployment automation tools, integration capabilities with decentralized storage, and real-time transparency assurance systems, enabling developers to build sophisticated decentralized AI applications without having to worry about the complex details of blockchain technology.

Heinrich clearly stated that “we plan to launch the mainnet by the end of the quarter.” This refers to the end of Q3 2025, meaning the end of September. The company is currently in final preparation stages for the mainnet launch, focusing on validator setup and core infrastructure arrangement.

The current testnet Galileo (V3 testnet) has recorded 339 million transactions, 6 million active accounts, 8,000 validator nodes, and peak throughput of 2,500 TPS (transactions per second). This testnet is the third-generation testnet launched on April 23, 2025, providing an environment where developers can build and deploy AI-specific applications without cost.

With the mainnet launch, full functionality including intelligent NFT transfers and agent deployment will be enabled. A Token Generation Event (TGE) is also planned for Q3, with users who set up and operated validator nodes during the testnet period receiving 1,143 tokens. Distribution will be one-third at TGE, with the remaining two-thirds vesting gradually over three years.

The Land of the Rising AI: Heinrich’s Japanese Gambit

Heinrich shows a positive attitude toward business expansion in Japan. 0G Labs has operated as a globally distributed organization from its inception, with employees placed not only at the San Francisco headquarters but around the world including China, Dubai, the UK, the United States, and Canada.

Leveraging this decentralized organizational structure, they aim to establish a local presence in Japan as well. As a specific expansion strategy, they plan to advance market development by first hiring an excellent Japanese GM (General Manager).

What’s interesting about the Japanese market is that, for example, you see Studio Ghibli images everywhere. This shows the characteristics of Japan’s legal environment. Basically, there’s a legal environment where you won’t be sued for intellectual property rights even if you use such copyrighted works for AI model training.

This means there’s a tolerant attitude toward open-source thinking from an intellectual property perspective. For us, who are advancing open-source AI model development, such an environment is very advantageous for business expansion, and we consider the Japanese market attractive.

However, interestingly, in the cryptocurrency field, it’s quite different. For example, comparing Korea as a market, 32% of all Koreans have cryptocurrency accounts, while in Japan it’s about 6 million people (about 4% of the total population).

So cryptocurrency adoption is significantly behind other regions of the world, but AI adoption and thinking are significantly ahead. These contrasting characteristics form an interesting market, and we consider entry into the Japanese market a long-term investment. (Heinrich)

The company is also advancing collaborations with Japanese investors. Japanese investors, including Gumi Cryptos Capital, a blockchain investment fund established by Japanese gaming company gumi, are participating in the company’s shareholder structure, establishing a foothold in the Japanese market. Additionally, while not yet officially announced, it was revealed that NTT Docomo is scheduled to participate as a validator in their network.

Beyond Human Limits: How AI Will Democratize Web3

Heinrich mentioned institutional-level adoption changes regarding cryptocurrency and blockchain proliferation. Before the Tokyo event, he had participated in the “Wyoming Blockchain Symposium” hosted by SALT and Kraken in Jackson Hole, Wyoming, where even Federal Reserve Board members were in attendance.

Why are companies and financial institutions starting to adopt cryptocurrency now? With traditional credit card payments, you need to pay 3% as transaction fees (author’s note: actual rates vary by individual contracts), but if you conduct completely on-chain payments on blockchain, you can achieve the same transaction for just a few cents or less,. (Heinrich)

While there are signs of hope for accelerated institutional adoption of Web3, significant challenges remain for general user adoption. Current Web3 technology’s technical complexity creates high barriers to entry for general users. However, Heinrich believes AI technology could automate these complex operations and dramatically improve usability.

AI can actually remove all the bad user experiences we’re familiar with from the crypto side—wallet creation, seed phrases, bridging between different chains. If AI agents can perform these 10 different steps in seconds, suddenly there’s potential for broad adoption of cryptocurrency and Web3 technology. (Heinrich)

Toward realizing this vision, 0G Labs is steadily advancing technological development. They’re currently at the stage of moving from testnet verification to full-scale service launch. The decentralized AI ecosystem envisioned by the company holds the potential to fundamentally transform the transparency and safety of AI when it functions as social infrastructure, going beyond mere technical improvements.

The company is also focusing on community building in the Japanese market, operating Japanese communities on Discord and X platforms. Showing strong enthusiasm for cooperation with Japanese companies, Heinrich states:

Please follow us on social media. We have Japanese communities on Discord and X, so please continue to engage with us through these channels. At the same time, we want to find really interesting use cases with different Japanese companies that make sense to have this intersection of Web3 and AI. We want to highlight actual use cases too, so we’re always open to collaboration from a business development perspective. (Heinrich)

0G Labs, advancing pioneering efforts in the decentralized AI field, presents solutions using blockchain technology to address the transparency and safety challenges of centralized AI. Heinrich’s expectations for the Japanese market and long-term vision herald the arrival of a new era where AI and blockchain converge.