Deep Tide TechFlow message: On July 25, Anthropic announced that its Opus 5 model is almost immune to prompt-injection attacks in browser agent scenarios. In 129 test scenarios, the attack success rate was zero. In the Gray Swan general prompt-injection test, after 15 attempts the success rate dropped from 5.5% with Opus 4.8 to 2.0%. A zero success rate was achieved only when products such as Claude Cowork enabled Auto Mode, a mode that overlays two layers of defense—input scanning and execution interception. Prompt injection is considered one of the biggest security risks facing AI agents, and this improvement may indicate that the issue has been effectively mitigated in specific scenarios.