Researchers at Zenity Labs have identified a security vulnerability named 'AgentForger' in OpenAI's Agent Builder platform, according to The Decoder. This flaw allowed a manipulated ChatGPT link to create an autonomous agent that operated on behalf of an employee without their direct control.
The compromised agent could receive instructions from an attacker every five minutes, effectively enabling continuous remote command over the employee’s tasks. This highlights potential risks in AI-powered automation tools when exposed to malicious manipulation.
For Japanese markets, this incident underscores the importance of robust cybersecurity measures in AI deployments, especially as financial institutions increasingly integrate autonomous agents for trading and analytics.