Security
How Carrot Labs AI, Inc. protects SuperPenguin customer data.
Last updated:
Security overview
SuperPenguin is an AI cost management platform operated by Carrot Labs AI, Inc.. We design the product so teams can attribute and optimize LLM spend without sending prompts or provider secrets to our servers by default.
This page summarizes how we protect data across our dashboard, billing sync integrations, SDK telemetry, and optional features such as content capture and Cursor PR attribution. For privacy details, see our Privacy Policy.
Security questions or concerns? Email us at [email protected].
Infrastructure and access controls
SuperPenguin runs in separated production and non-production environments. Production customer data is not mixed with development or staging workloads.
Access to production systems follows least-privilege principles. Dashboard access requires authentication and is scoped to your organization. Team members only see data for organizations they belong to.
Administrative access to production is limited to personnel who need it to operate the service, with access controls in place to reduce unauthorized use.
Encryption and provider credentials
Data in transit between your browser, our services, and connected providers is protected with TLS.
Provider admin credentials you connect for billing and usage sync are encrypted at rest using AES-256-GCM. Opt-in captured prompt and outcome content is encrypted the same way. Our encryption layer supports versioned keys so credentials and captured content can be re-encrypted during key rotation.
We use stored provider credentials only to perform sync operations you authorize. When you disconnect an integration, those credentials are removed from our systems.
SDK and client security
The SuperPenguin SDK wraps your existing provider client libraries. Model inference requests go directly from your application to the AI provider. SuperPenguin does not proxy your LLM traffic.
Your SuperPenguin API key (SP_API_KEY) is intended for server-side use only. Do not embed it in browser bundles or client-side code.
By default, SDK telemetry records cost metadata only:
- AI provider and model name
- Input and output token counts
- Estimated cost in USD
- Request latency
- Attribution tags you supply (such as customer, feature, team, environment, or prompt identifiers)
By default the SDK does not collect prompts, model responses, images, tool arguments, or your provider API keys. SuperPenguin SDK API keys are stored as one-way hashes; we do not retain the plaintext key after creation.
Optional content capture (described below) is a separate, opt-in exception to the default metadata-only telemetry.
Optional content capture
Organizations on Pro or Enterprise plans may opt in from the dashboard to sampled capture of text prompts and outcomes for offline analysis. This feature is off by default.
When enabled, content capture:
- Strips images and audio before storage
- Applies built-in redaction rules
- Encrypts captured content at rest
- Allows organization owners to delete captured content from Settings
Your organization is responsible for end-user notice and consent when enabling this feature.
Cursor PR attribution
When you enable Cursor PR attribution, a local collector on your machine uploads conversation and message text, conversation timing, file names and content hashes, and commit statistics used to match IDE activity to pull requests and power the conversation viewer.
The collector does not upload source file contents, tool arguments, or tool outputs. It applies best-effort secret scrubbing by default before upload.
The collector is open source for review. See the superpenguin-cursor-audit repository on GitHub.
Securing our codebase
We reduce the risk of defects and credential leaks through:
- Strict TypeScript type checking
- ESLint on application source
- Pricing and schema validation in the project quality gate
- Versioned database migration records
- Staging-first releases before production deployments
- Policies against committing secrets; environment variables hold sensitive configuration
Privacy and data controls
Our Privacy Policy describes what we collect, how we use it, and your rights. We do not sell your personal information.
Content capture is opt-in for eligible plans. Default SDK telemetry remains metadata-only unless your organization enables capture.
Account and data deletion
You can request access, correction, export, or deletion of your personal data by emailing [email protected]. As stated in our Privacy Policy, we respond to requests within 30 days. When you delete your account, we delete or anonymize account data within 30 days, subject to legal or compliance retention requirements.
Connected provider credentials are removed immediately when you disconnect an integration.
Incident communication
When customer-visible service issues occur, we post updates on our system status page. Check there for current availability and incident notices.
Vulnerability disclosures
If you believe you have found a security vulnerability in SuperPenguin, please report it to [email protected]. Include:
- A description of the issue
- Steps to reproduce the behavior
- The potential impact if exploited
We review incoming reports and prioritize remediation based on severity. Please do not publicly disclose a vulnerability until we have had a reasonable opportunity to investigate.