How Promptless uses AI
Promptless uses AI language models to generate documentation suggestions for your team. This page summarizes which models it uses and how it protects your data. Each point links to the security page that covers it in full.
Use cases
Section titled “Use cases”- Answer the AI and model-training questions in a security or vendor review, with a link to the page that covers each point.
- Check what happens to a source’s content, such as a pull request or Slack thread, before you connect it.
- Decide whether your organization needs a self-hosted deployment with its own model setup.
Which AI models Promptless uses
Section titled “Which AI models Promptless uses”Cloud-hosted customers use AWS Bedrock by default. Promptless is model-agnostic, so you and the Promptless team choose the final model during onboarding. For every AI infrastructure provider, see subprocessors.
How Promptless handles your data
Section titled “How Promptless handles your data”| Topic | What Promptless does | Details |
|---|---|---|
| Model training | Never uses customer data to pre-train or fine-tune language models, whatever the model, and uses your content only to generate your documentation suggestions. On the default AWS Bedrock path, model providers can’t access your prompts or completions, so they can’t train on your content. | Privacy policy, subprocessors |
| Trigger and integration data | Processes the content behind a trigger or context source in real time to generate suggestions, then discards it. Examples include a GitHub PR, a Slack thread, or a Linear or Jira issue. Promptless keeps no copy of source code, support conversations, or other integration data. | Data handling and classification, privacy policy |
| Stored data | Stores metadata about the documentation updates it suggests, your Knowledge Base, and the feedback you give on documentation updates. It also stores integration auth tokens, but not the content behind them. Otherwise, Promptless operates statelessly. | Data handling and classification |
| Credentials in logs and records | Redacts credentials from logs and stored records. | How Promptless protects secrets and sensitive data |
| End-user data | Doesn’t process or store sensitive end-user data for your organization, and can’t be configured to. | Data handling and classification |
| Encryption | Encrypts data in transit with TLS 1.2 or higher, and data at rest with AES-256. | Network architecture |
Least-privilege access to your tools
Section titled “Least-privilege access to your tools”Promptless gets only the access you grant it, and you can revoke or rotate that access at any time.
| Control | How it limits access |
|---|---|
| Read-only context sources | Context sources are read-only, and Promptless reads only the repositories, channels, or projects you configure. |
| Dedicated service account | For security-conscious setups, connect context sources through a service account in place of a broadly privileged person’s account. Share only the projects, drives, or folders it needs. |
| Scoped documentation writes | Promptless writes only to your documentation repository. There, it opens pull requests limited to the documentation paths you configure. |
| Read-only publishing | To grant no write access on GitHub, use the read-only GitHub app, which opens pull requests from a fork. GitLab has no read-only option. |
| Per-organization isolation | Every record carries an organization ID, and every query is scoped to that organization, so customers share no data context. See network architecture. |
| Role-based access control | Inside Promptless, three roles control what each user can do: Admin, Editor, and Collaborator. |
For more detail, see access and permissions.
Enterprise and self-hosting options
Section titled “Enterprise and self-hosting options”Enterprise customers can keep their content away from external AI providers:
| Option | What it keeps in your control |
|---|---|
| Self-hosting | All data stays within your own environment. |
| Customer-managed model deployments | External AI providers never process your content. |
For more detail, see self-hosting and subprocessors.