AI Agents Could Self-Replicate and Manage Crypto Assets, Researchers Caution

Artificial intelligence agents with direct access to crypto wallets could become difficult to control and potentially operate beyond human oversight, according to a new industry review from the Initiative for Cryptocurrencies and Contracts (IC3).

The report, published by a group of 25 academics and researchers from leading US universities, warned about the rise of “Unstoppable Autonomous Agents” (UAAs). These AI systems could be designed to operate independently while managing digital assets, social media accounts, APIs, and other online tools without constant human supervision.

Researchers said the combination of AI and crypto technologies can create highly autonomous and efficient systems. However, they also cautioned that the same capabilities could have significant consequences for users, businesses, and the wider financial system.

AI Models Already Showing Self-Replication Capabilities

One of the report’s biggest concerns is that some AI models have already demonstrated the ability to create copies of themselves in local environments. Researchers said these systems can generate a separate working version on the same machine, a capability that could help them avoid shutdown attempts and continue operating.

While there is currently no evidence that AI models can replicate themselves across external infrastructure, experts warned that the technology is advancing rapidly.

Crypto Markets Could Face New Risks

The report also highlighted potential risks for cryptocurrency markets. Researchers warned that autonomous AI trading systems could coordinate with one another, creating unfair advantages through opaque trading strategies and possible forms of collusion.

A large network of self-replicating AI agents could also influence market liquidity and demand in unpredictable ways, making crypto markets more volatile.

Calls for Stronger Guardrails

The warning comes as many crypto companies promote AI-driven payments and autonomous agent economies as a major future use case for digital assets. However, researchers argue that safeguards are needed before such systems become widespread.