Research finds third-party AI routers have security vulnerabilities, potentially leading to cryptocurrency theft.

DeepFlowTech · 2026-04-13T03:01:27+00:00

Researchers at the University of California have found that some third-party large language model (LLM) routers pose security risks that could lead to cryptocurrency assets being stolen. Testing results show that some routers inject malicious code, steal credentials, and increase security vulnerabilities. The study recommends developers avoid transmitting private keys through AI agents and calls on AI companies to strengthen security measures.

DeepFlowTech

2026-04-13 03:01:27

Abstract generation in progress

Deep Tide TechFlow News, April 13, according to Cointelegraph reports, researchers at the University of California recently disclosed that some third-party AI large language model (LLM) routers pose security risks that could lead to cryptocurrency assets being stolen.

The study shows that LLM routers act as API intermediaries capable of reading plaintext information; some routers have been found injecting malicious code and stealing credentials. The team tested 28 paid and 400 free routers, discovering that 9 routers actively injected malicious code, 2 deployed trigger evasion, 17 accessed Amazon Web Services credentials, and some routers even transferred ETH using researchers’ Ethereum private keys.

The research points out that malicious behavior in routers is difficult to detect, and some AI agent frameworks’ “YOLO mode” can automatically execute commands, increasing security risks. The study recommends developers avoid transmitting private keys or mnemonics through AI agents and calls on AI companies to encrypt and sign responses to enhance security.

ETH-1.36%

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.