Cross-request KV cache exposure in the GLM-5.3 Workers AI model, and our remediation with SGLang
A configuration mismatch in GLM 5.3 on Workers AI triggered an SGLang bug that exposed KV cache pages across requests. Here is how we addressed the issue.
Read the original at Cloudflare Blog ↗
- Topics
- Endpoint & Network Security
- Related
- Adapting to AI penetration testing orchestrators: The Modern Security Model Cyber Risks · Nov 13, 2026Introducing Spend Management: Spending limits and alerts for Cloudflare Cyber Risks · Nov 13, 2026Choosing Que Paxa, an unusual consensus algorithm Cyber Risks · Nov 11, 2026