Experts Warn of Security Risks in OpenAI’s New ChatGPT Atlas Browser

OpenAI recently launched its new AI-powered web browser, ChatGPT Atlas. The AI browser s currently available for macOS, with Android, iOS, and Windows versions on the way. Not to mention, the launch caught eyeballs of many, as Atlas looks to transform browsing by letting users search, plan, and even book services directly through ChatGPT.

However, cybersecurity experts are already raising red flags over potential vulnerabilities in Atlas (via Fortune). They warn that AI browsers like Atlas could be exposed to prompt injection attacks, where malicious webpages feed hidden instructions to the AI to extract data or perform harmful actions.

These hidden prompts could trick the AI into revealing personal information or accessing sensitive accounts without user intent. “There will always be some residual risks around prompt injections because that’s just the nature of systems that interpret natural language,” said George Chalhoub, assistant professor at the UCL Interaction Centre.

Responding to concerns, Dane Stuckey, OpenAI’s Chief Information Security Officer, said the company is “very thoughtfully researching and mitigating” these risks. “Our long-term goal is that you should be able to trust ChatGPT agent to use your browser, the same way you’d trust your most competent, trustworthy, and security-aware colleague or friend,” Stuckey added.

He added that OpenAI has introduced overlapping guardrails, red-teaming, and rapid response systems but acknowledged that “prompt injection remains a frontier, unsolved security problem.”

Have you tried the Atlas browser yet? If you haven’t and looking to download and learn how tio use it, check our guide.