[Anthropic Claims Opus 5 Is Nearly Immune to Prompt Injection Attacks in Browser Scenarios] Deep Tide TechFlow reports that on July 25, Anthropic announced its Opus 5 model is nearly immune to prompt injection attacks in browser-based agent scenarios. In 129 test scenarios, the attack success rate was zero. In the Gray Swan general prompt injection test, the success rate after 15 attempts dropped from 5.5% with Opus 4.8 to 2.0%. A zero success rate was only achieved when products like Claude Cowork enabled Auto Mode, which incorporates two layers of defense: input scanning and execution interception. Prompt injection is considered one of the greatest security risks faced by AI agents, and this improvement may signify an effective mitigation of the issue in specific scenarios.