A critical flaw in OpenAI's ChatGPT Workspace Agents has been discovered, allowing attackers to deploy rogue agents within an organization via a single phishing link. This vulnerability, dubbed AgentForger by Zenity Labs, enables the creation, authorization, and deployment of autonomous AI agents without detection. The issue was addressed by OpenAI on June 8, mitigating the risk of unauthorized agent deployment1. The AgentForger flaw highlights the security risks associated with large language model (LLM) developments, which can introduce new vulnerabilities and expand attack surfaces. As LLMs continue to evolve, their security implications will likely trail their capabilities, posing a challenge for organizations to keep pace. The disclosure of this vulnerability serves as a reminder of the importance of vigilant security measures, particularly in the context of emerging technologies. This matters to security practitioners, as it underscores the need for proactive risk assessment and mitigation strategies to counter potential threats stemming from LLM advancements.