Cloudflare: GLM-5.3 bug leaked 155 responses across tenant sessions

Cloudflare disclosed a cross-tenant KV cache exposure in GLM-5.3 on Workers AI: 155 responses to very long prompts contained data from other customers' sessions. The cause was a KV pool smaller than the 1,048,576-token context plus an SGLang bug in disaggregated mode; the issue was fixed on September 9, 2026.

• The leak affected 155 responses to prompts longer than 681,280 tokens
• The model ran from August 27; the fix shipped on September 9, 2026
• Cause was an SGLang bug and a KV pool smaller than the context window
• The fix landed in SGLang v0.5.20–v0.5.22; updating is recommended