Opus 5 may have solved browser-based prompt injection, the biggest security flaw haunting AI agents

Opus 5, combined with Auto Mode, achieves a zero percent prompt injection success rate for browser agents across 129 test scenarios. Without these additional protection layers, the success rate stands at 3.7 percent. If these numbers hold true in practice, Anthropic may have addressed one of the most significant security challenges facing AI agents operating in browsers. This advancement marks a crucial step towards enhancing the security of AI agents in browser environments, potentially impacting how these systems are utilized in the future.