GCP Model Armor Enhances AI Prompt and Response Screening with Larger Token Support

GCP's Model Armor feature now supports screening AI prompts and responses up to 65,536 tokens, equating to 262,144 characters. This expansion significantly boosts its capacity for prompt injection and jailbreak detection, responsible AI compliance, and child sexual abuse material (CSAM) filtering. The capability is crucial for developers working with large language models, ensuring robust content moderation and safety. Additionally, Model Armor scans the first 256 URLs found within the screened prompts and responses.