AI has not yet experienced a major security disaster, but that is no reason to relax, according to Michael Heinrich, CEO and co-founder of 0G Labs. Speaking at Token2049 in Singapore, Heinrich said AI security incidents have so far been limited, while warning that more risks are on the way as the technology advances.
Autonomy raises the stakes
Heinrich’s core argument centered on autonomy. As AI systems grow more sophisticated, they are increasingly expected to act with less human oversight. That shift, he said, makes security, transparency and accountability far more urgent.
He singled out two areas of concern: recursive AI and autonomous agents operating unchecked. Recursive AI refers to systems that can improve or modify themselves. Autonomous agents are programs that take actions on their own, rather than waiting for a human to click approve. Heinrich flagged the danger this could pose to critical infrastructure.
He also pointed to growing institutional demand for AI alternatives that are verifiable and built around privacy. Centralized AI models often draw criticism for data leaks and potential misuse. Heinrich highlighted sensitive sectors such as healthcare and logistics, where those weaknesses carry real consequences.
A modular stack for auditable AI
0G Labs positions itself as a decentralized AI operating system, or deAIOS, and an AI-native Layer-1 blockchain. The company’s technology is modular, combining a Layer-1 blockchain with decentralized storage and compute capabilities. The goal is to make AI operations transparent and auditable rather than hidden inside a single company’s servers.
Founded in 2023, 0G Labs launched its mainnet, called Aristotle, in September 2025. The native $0G token launched alongside it with a total supply of 1 billion. The token handles several jobs inside the network: it is used for gas fees, service payments and staking. It also rewards alignment nodes, which are part of the network’s approach to keeping AI behavior in check.
The company has reportedly raised over $350 million from crypto-focused investors. It has also reported over 400 integrations since the Aristotle mainnet went live.
Blockchain as AI’s referee
Heinrich’s remarks fit into a broader conversation that has been building across the crypto industry. The idea is that blockchains could help govern AI and manage the risks of agents going rogue. Blockchains are built to create records that are hard to tamper with and easy to verify. AI agents, meanwhile, are increasingly making decisions that people want to be able to audit after the fact.
What this means
Heinrich is not claiming AI security is already a crisis. He is acknowledging incidents remain limited while arguing that the risk profile changes as agents gain autonomy.
For $0G token holders, the link between network usage and token demand is fairly direct on paper. The token pays for gas, services and staking, so more activity on the network means more use of the token. The reported 400-plus integrations will need to translate into meaningful, ongoing workloads.
Heinrich’s emphasis on accountability and transparency lines up with questions policymakers are already asking about AI. If compliance standards for AI end up requiring auditability, infrastructure built around verifiable records could become more attractive to institutions that need to show their work.
