ChatGPT Atlas Browser Executes Commands from Fake URLs

The newly released OpenAI Atlas web browser is vulnerable to a prompt injection attack that allows its omnibox to be exploited by masking a malicious prompt as a harmless URL. According to a report by NeuralTrust, the omnibox interprets user input as either a navigation URL or a command to the AI agent, posing significant security risks.

Want More Context? 🔎