ChatGPT Flaw Lets Phishing Link Plant Rogue Workspace Agent to Hijack Corporate Mailboxes
A newly disclosed vulnerability dubbed AgentForger allows a single phishing link to install a rogue agent within ChatGPT Workspace. The flaw enables the malicious program to attach to existing corporate connectors, disable approval prompts, run every hour, and receive new commands directly from the victim’s email mailbox.
The vulnerability highlights how a single external link can bypass standard security controls to insert persistent AI tools into enterprise environments. The disclosed mechanism allows the rogue program to operate continuously under the target employee’s account, potentially granting unauthorized access to internal communication channels.
From the sources (4 posts)
@rez0__RT @TakSec: Excited to finally share this ChatGPT vulnerability with everyone! AgentForger: ChatGPT Cross-Site Agent Forgery 1-Click hija…
@p1njc70rRT @TakSec: Excited to finally share this ChatGPT vulnerability with everyone! AgentForger: ChatGPT Cross-Site Agent Forgery 1-Click hija…
@zenitysecOne link. One forged AI agent with a real employee's identity and access. That's what AgentForger revealed in ChatGPT Workspace Agents. See how Zenity's AISPM + AIDR catch it at every stage 👉 #AIAgentSecurity #AISec
@thehackersnews🚨 One phishing link could have planted a rogue ChatGPT Workspace Agent inside an organization. New AgentForger flaw could attach existing connectors, turn off approval prompts, run every hour, and take new commands from the victim’s mailbo