Researchers at Anthropic have been studying how AI agents interact with CAPTCHAs—the security tests designed to verify that a user is human. Their findings suggest that these artificial intelligence systems experience difficulties with the tests, much like many people do.

The research offers a glimpse into how AI agents approach problems when they encounter common internet security barriers. As AI systems become more capable and autonomous, understanding how they handle obstacles like CAPTCHAs becomes increasingly relevant for both security and AI development.