Cloudflare reports cross-request data exposure in a Workers AI model
Facts read 16 Nov 2026. Page reviewed 16 Nov 2026.
Cloudflare's 26 October 2026 post says a configuration mismatch in one hosted model exposed data across customers' requests between 27 August and 9 September 2026. The company states 155 responses were affected and that it found no evidence of deliberate exploitation.
Cloudflare states that a small number of customers using the GLM 5.3 model on Workers AI received responses containing data from other customers' requests made shortly before theirs, all for prompts longer than 681,280 tokens.
The post says the issue was reported on 8 September through its bug bounty program, reproduced the next morning, and the model was taken offline within 40 minutes of declaring an incident and restored on 9 September. It says no other Workers AI model was affected.
Cloudflare published the report alongside a post from the maintainers of the open source inference engine it uses. The ledger would class this as a cross-tenant exposure at the provider.
Related pages
Sources
- https://blog.cloudflare.com/kv-cache-exposure-remediation/, read 16 Nov 2026
Last reviewed 16 Nov 2026.