Autonomous AI Agents Create Crypto Financial Risks

[ad_1]

Artificial intelligence agents with access to autonomous crypto wallets could become unstoppable if they are deployed maliciously or escape from sandboxes, experts at an academic research consortium have warned.

According to Industry on June 8, “unstoppable autonomous agents” (UAAs) pose a clear threat if deployed to continue live and access digital assets. Evaluation Written by 25 academics and experts from top US universities for the Initiative on Cryptocurrencies and Contracts (IC3).

“When integrated into a system, crypto tools can channel the fluid power of AI into secure, reliable, and highly autonomous systems,” the researchers wrote. However, this combination “could have many consequences for consumers and the financial system,” he said.

UAs may have access to cryptocurrency wallets, social media accounts, APIs and other external tools, the researchers said.

“The capabilities to enable such agents are already emerging and evolving rapidly.”

The warning comes as there were crypto projects and executives Pushing This year, the narrative of the agent payment and micropayment economy suggests that this could be the biggest advantage of decentralized digital assets.

Table of Contents

AI self-replicating alarm bells

The paper also notes that existing models can already “exceed the red lines of self-replication,” by autonomously creating copies of themselves on the same machine, creating “the potential for system closure and scalability.”

Because reward cues used in training often fail to capture the intended goals, “UAEs deployed for good intentions can inadvertently cause harm” or pursue resource acquisition as a default strategy, he said.

However, the authors point out that the models have not yet replicated themselves to external infrastructures.

Potential AI agent internal business benefits

Self-replicating, resource-generating agents can create unpredictable demand and volatility in crypto markets.

“AI-powered trading systems may allow cooperation between autonomous agents and create unfair insider advantages through opaque strategies.”

Related: China already has its own Mythos-like AI: Nvidia CEO to train

The tech sector is already dealing with tough questions about the unmitigated threat of AI.

Models such as the Anthropic Cloud Mythos have appeared before. able Finding and exploiting zero-day vulnerabilities in major operating systems.

Meanwhile, Gartner warned In late May, it predicted that governance failures around autonomous AI agents could lead to widespread corporate failures, with 40% of companies forced to lay off agents by 2027.

“The damage from such fully autonomous agents is severe,” the researchers said, suggesting circuit-breaking protection mechanisms.