Cloudflare tested their WAF with frontier AI models, finding that it blocked 99.9% of attacks, with the remaining 0.1% requiring manual review. They share their approach and provide guidance on how to correctly deploy WAFs and patch software.