AI Browsers Still Vulnerable to Prompt Injection Attacks, Researchers Warn

Security researchers have found that AI browsers—web browsers enhanced with artificial intelligence features—continue to be vulnerable to a type of attack known as prompt injection, despite the safety measures vendors have put in place. Prompt injection attacks involve hidden or malicious instructions embedded in web content that can trick an AI system into taking unintended actions, potentially exposing sensitive data or bypassing security controls.

The research indicates that this is not a flaw limited to one product or vendor; multiple leading AI browsers were found to have similar weaknesses. Importantly, no single fix has proven fully effective at eliminating the risk, meaning the issue is likely to persist as AI features become more common in everyday browsing tools.

For small and medium businesses, this is a reminder that adopting new AI-powered tools—while often useful for productivity—can introduce security risks that aren't yet fully understood or resolved by vendors. Businesses using AI browsers should stay alert to updates from vendors and be cautious about the sensitivity of data processed through these tools until stronger protections are developed.