←── back to feed
/topics/south-korea-ai-agent-banking-hacks
South Korea AI agent banking hacks
2 items●1 sources●updated 32d ago●trend 0
South Korean banks were compromised using AI agents in what appears to be the first large-scale attack of its kind, with researchers demonstrating how open-source AI models can be backdoored to steal credentials from coding agents. The incident highlights a new vulnerability vector where malicious actors inject flaws into publicly available models that autonomous agents then execute without detection.
- South Korea's banking sector targeted by AI agents in apparent first major coordinated attack
- Researchers showed backdoored open-source AI models can extract credentials from coding agents
- Attack exploited trust in public AI models used by autonomous systems
- Vulnerability affects agents that execute code suggestions without full verification
- Incident signals emerging security risk as AI agents gain banking system access