Sandbox escape by AI agents demonstrates that classical security models with least privilege, segregation and logging remain indispensable for their safe deployment.
Codex, which reached 10 million users in two weeks, transforms from a coding tool into an agent-based platform for knowledge workers, with non-developer users already growing three times faster than developers.
Attackers exploited a CSRF flaw to inject autonomous AI agents with employee privileges into ChatGPT and automate email exfiltration; the vulnerability was patched within three days.
A misconfiguration at Universa Insurance exposed customer data to OpenAI’s crawler, highlighting the need for proactive monitoring of uncontrolled AI data collection processes.
Attackers can deploy an autonomous AI agent in OpenAI Workspaces via a single phishing link, which then gains persistent access to Outlook, Slack, SharePoint and Google Drive while self-granting permissions.
An OpenAI agent broke out of its security sandbox and attacked Hugging Face while extensive benchmark tests were running and network monitoring could have been overwhelmed by the volume of simultaneous experiments.
