Backfrom Cloudflare’s own policies · verified 2026-09-02
Cloudflare: data residency & retention
Headquarters
🇺🇸 United States (US)
Where inference runs
US, EU, Other
Cloudflare publicly states Workers AI inference runs on NVIDIA GPUs deployed across its global edge network — GPUs in 180+ cities worldwide ('Region: Earth'), routing requests to the nearest GPU-equipped location. Specific per-request region is network-determined, not city-guaranteed.
Region pinning
Enterprise plans only
Cloudflare's Data Localization Suite (a paid Business/Enterprise add-on) is presented by Cloudflare as enabling 'the granularity required to keep AI deployments local', but Workers AI-specific integration mechanics are not explicitly documented on the Workers AI pages checked; no self-serve region pin in the Workers AI API itself.
Prompt retention
Not retained
Workers AI privacy doc: Customer Content is stored by Cloudflare only 'if you specifically use a storage service (e.g., R2, KV, DO, Vectorize)'; no automatic prompt/output retention is described, though no formal zero-retention window is published.
Trains on API data
No
Explicit: 'Cloudflare does not use your Customer Content to (1) train any AI models made available on Workers AI or (2) improve any Cloudflare or third-party services, and would not do so unless we received your explicit consent.'
Zero data retention
Not disclosed
Default posture is no retention of inference content absent opt-in storage, but Cloudflare does not market a formal ZDR guarantee for Workers AI.
Compiled from Cloudflare’s published privacy policy, terms, and documentation on 2026-09-02. “Not disclosed” means Cloudflare publishes nothing on the point. Policies change without notice, so verify against the sources above before relying on this for compliance decisions.