OpenAI’s new Atlas browser could have extremely concerning security issues, experts warn – here’s what we know


  • All agentic AI browsers are sensitive to indirect prompt injections
  • Only use agent navigation when you are not handling sensitive information
  • We may need to rethink how browsers work and how we use them

Just days after OpenAI released Atlas, its version of the web browser, the company is struggling to maintain its reputation in the face of security concerns.

The Chromium-based browser, which has a built-in AI agent for web navigation and automation, was found to be vulnerable to indirect prompt injection, meaning malicious commands can be hidden in web content to manipulate the agent’s functionality.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top