Data retention policy
Swa retains personal data only as long as necessary to fulfill the requested functionality or to meet legal obligations, such as payment processing.
Swa retains prompt inputs and LLM outputs for the purpose of providing context, continuity, and fast retrieval. This data is securely stored in a vector database using retrieval‑augmented generation (RAG). Retention is limited to the duration of Swa’s active presence in a channel or workspace. Data from a Channel is automatically removed from Swa when Swa is removed from the Channel, and no long‑term storage occurs outside of these contexts.
Operational logs (e.g., API call metadata) are retained temporarily and are deleted or anonymized when no longer needed.
Usage logs are captured for billing and audit purposes which includes the model used and token count sent and received per transaction.
Data archiving and removal policy
For connected data sources like Slack, any contextual data is deleted
Users may remove stored integration data at any time via admin panels, admin channels, and requests.
If Swa is uninstalled from a Slack workspace, all associated workspace data will be automatically deleted within 72 hours. Once deleted, the data is permanently unretrievable. Similarly, if Swa is removed from a channel, all related channel context is purged automatically.
Swa will comply with all additional deletion requests within a reasonable timeframe - typically immediately. Email support@swa-ai.com for further questions or requests!
Data storage policy
All user data is stored and processed within U.S.-based infrastructure. OAuth tokens and credentials are secured in an encrypted environment using industry‑standard encryption protocols (including TLS, HTTPS, and OAuth 2.0).
Swa does not use customer data for model training. External content is never stored or replicated unless explicitly required by the user - for example, when provided as part of a file upload or context‑enhanced task.
Data center location(s)
United States
Data hosting details
AWS, Azure, & Google Cloud, Pinecone
Data hosting company
AWS, Azure, GCP, Pinecone
App/service has sub-processors
no
App/service uses large language models (LLM)
yes
LLM model(s) used
ChatGPT, Claude, Perplexity, Deepseek, Llama, and others.
LLM retention settings
Swa does not persist messages or prompts sent to LLMs. Outputs may be temporarily cached in-memory for reply chaining (max a few seconds). No training or fine-tuning is performed on user data. LLM Providers used do not retain or train data.
LLM data tenancy policy
Swa uses third-party LLM APIs (e.g. OpenAI, Anthropic, Perplexity, Grok) in a fire-and-forget model. Inputs are sent to the LLM provider for inference only; no data is stored beyond processing.
LLM data residency policy
Inference requests may be processed in the LLM provider’s default region (typically US-based). Swa does not control regional residency unless the provider offers and enforces such controls.