Llama
GDPR Compliance
Data Handling
Meta’s publicly available processor terms state sub-processors may be based in the United States, the EU/EEA, or other countries. I could not find a primary-source statement specific to the Llama API guaranteeing EU-only processing or confirming where inference GPUs are located, so EU-only processing and inference locality are not confirmed.
A Llama developer help page about whether Meta uses Llama API data to train models exists, but its contents were not publicly accessible without login from the available tooling, so I could not verify the actual policy text from a primary source.
Certifications & EU AI Act
Meta publishes EU-specific AI terms for its AI services, but I could not find a primary-source statement in the consulted sources claiming formal EU AI Act compliance for the Llama API specifically.
Verification
- https://llama.developer.meta.com/docs ↗
- https://llama.developer.meta.com/help/policy/does-meta-use-my-llama-api-data-to-train-its-models ↗
- https://llama.developer.meta.com/docs/trust/data-commitments ↗
- https://about.fb.com/br/company%20info/ ↗
- https://about.fb.com/news/2017/12/moving-to-a-local-selling-model/ ↗
- https://www.facebook.com/legal/terms/dataprocessing ↗
- https://www.facebook.com/legal/terms/Privacy/GDTA ↗
- https://www.facebook.com/privacy/policy/ ↗
- https://www.facebook.com/legal/eu-ai-terms ↗
Meta/Llama offers a public data processing terms page with SCC-based transfer addenda and references to a sub-processor list, and those terms mention annual SOC 2 Type II audit reporting. However, several Llama API-specific trust pages were not publicly accessible without login in the available tooling, so Llama-specific training, retention, and EU inference-location facts could not be confirmed and are left null.