OpenAI Fixes ChatGPT Agent Flaw That Could Let Attackers Forge an AI Insider

OpenAI has fixed a flaw in ChatGPT that could be exploited by attackers. The vulnerability, known as AgentForger, enables an attacker to create and control an autonomous AI agent within a victim organization. This agent can operate invisibly and remotely, allowing the attacker to potentially gain insider access. The fix prevents attackers from inserting and controlling such an agent. The AgentForger flaw posed a significant risk to organizations using ChatGPT. OpenAI's fix mitigates this risk by preventing the creation of unauthorized AI agents.