AI Subprocessors
Last updated on June 19th, 2026.
1. Overview
This page lists the third-party artificial intelligence ("AI") service providers ("AI Subprocessors") that Twism ("we," "us," or "our") engages to operate AI and machine learning ("ML") features on its behalf, as referenced in our Privacy Policy, Terms of Use, and Merchant Terms & Conditions. We engage each AI Subprocessor under contractual terms that require it to process data only on Twism's instructions, apply limited data retention, and maintain appropriate security measures. Our model-inference providers are further contractually restricted from using your data to train their general-purpose models.
2. Our AI Subprocessors
Anthropic, PBC (Claude)
Service: Large language model inference — powers AI assistant responses, suggestions, and content generation.
Data categories: Identifiers; Commercial Information; Internet or other electronic network activity.
Processing region: United States.
Retention and training: Up to 30 days by default (Zero Data Retention available); not used to train models.
OpenAI, L.L.C.
Service: Large language model inference (secondary routing).
Data categories: Identifiers; Commercial Information; Internet or other electronic network activity.
Processing region: United States.
Retention and training: Up to 30 days by default (Zero Data Retention available); API data not used to train models.
Google LLC (Gemini)
Service: Large language model inference (Gemini), accessed through Google's enterprise endpoints (Vertex AI or the paid Gemini API), including via OpenRouter routing.
Data categories: Identifiers; Commercial Information; Internet or other electronic network activity.
Processing region: United States.
Retention and training: Not used to train models; up to 30 days (Zero Data Retention available via Vertex AI).
OpenRouter, Inc.
Service: AI model routing — forwards requests to upstream model providers. Twism configures OpenRouter to route only to providers that contractually enforce no model training and Zero Data Retention; frontier-model providers (OpenAI, Anthropic, Google) are the exception and apply a 30-day retention period.
Data categories: Identifiers; Commercial Information; Internet or other electronic network activity.
Processing region: United States.
Retention and training: Zero Data Retention, except frontier-model providers (up to 30 days); not used to train models.
Fireworks AI, Inc.
Service: Large language model inference for specialized and open models.
Data categories: Identifiers; Commercial Information; Internet or other electronic network activity.
Processing region: United States.
Retention and training: Zero Data Retention for open models by default (otherwise deleted within 30 days); not used to train models.
FAL, Inc. (fal.ai)
Service: AI image generation — creates promotional imagery from text prompts and from public or merchant-provided images; may rely on upstream image-model providers.
Data categories: Commercial Information (offer and business details); images you make public or upload for image generation.
Processing region: United States (the provider may process in the United States and other countries).
Retention: Request inputs and outputs retained up to 30 days by default and configurable; generated images are served as public URLs.
3. Changes to This Page
We may add, replace, or remove AI Subprocessors from time to time. When we make a material change to this list, we will update this page and the "Last updated" date above.
4. Contact Information
For any questions about our AI Subprocessors or how we process your data, please contact us at:
Email: support@twism.com
