26 Jun 2026 18:01 UTC
Fernando Irarrázaval posted his OpenClaw assistant's inbox to Hacker News and watched Claude Opus 4.6 hold off thousands of attackers.
➤ An AI assistant named Fiu, running on the OpenClaw framework with Claude Opus 4.6, successfully defended against over 6,000 prompt injection attempts via email.
➤ The attacks, designed to trick the AI into leaking sensitive information, were largely unsuccessful, though they caused issues with Gmail's fraud detection and increased API costs.
➤ The article highlights the effectiveness of the AI's security prompt against sophisticated attacks, contrasting it with the higher success rates of similar attacks on less robust AI models.
