Paid models use zero data retention; free tier models may log and retain data to improve the models.
AWS Bedrock does not use customer data (inputs, outputs, or custom model weights) to train or improve any base models by default. Other AWS AI services may collect data by default but allow account-level opt-out.
Does not train models on private user code. Zero Data Retention is active by default.
Consumer chats on Claude Free, Pro, and Max require user opt-in for model training. Claude Code sessions follow account settings. Console API and Claude for Work/Enterprise plans are excluded from training.
Groq does not train on customer data. Prompts and responses are retained for up to 7 days for abuse monitoring, then permanently deleted.
Cerebras does not harvest data from AI interactions. Prompts and outputs are not used for training purposes.
Fireworks AI does not use customer data to train models without explicit opt-in. Zero Data Retention is available for open models by default.
LiteLLM is an open-source proxy. No data training occurs - it simply routes requests to model providers.
Prompts and inputs are processed in real time and not saved on the servers for model training.