MisbehaviorSerious1 sourceFirst seen
GLM 5.3 on Workers AI handed some customers other customers' cached data
Ourtake
The exposure was present from launch and was found only after an outside researcher reported it.
What happened
Between August 27 and September 9, 2026, a small number of customers using Cloudflare's GLM 5.3 model on Workers AI received responses containing data from other customers' requests. 155 responses were affected, all to prompts longer than 681,280 tokens. The cause was a KV cache pool smaller than the context window, combined with a bug in the SGLang inference engine. Cloudflare disabled the model, fixed the configuration and engine, and restored it on September 9.