AI Data Use
Questions: [email protected].
Short Version
Enji Guard does not train or operate its own foundation models.
Enji Guard uses approved third-party AI model providers so the Enji Fleet agents orchestrator can run selected audits, reports, autofixes, scheduled improvement jobs, and other agent tasks.
Customer code and task data should be sent only to approved providers and routes with documented public data-use and retention terms. Do not describe all provider routes as zero-retention: some approved routes may retain limited abuse-monitoring logs or other technical state according to the provider’s published terms.
What Data May Be Sent To AI Providers
Depending on the selected action, task context may include:
- task title, task description, and runbook instructions;
- repository names, file paths, code snippets, diffs, logs, and test output;
- website URLs, HTTP responses, screenshots, or reproduction steps from authorized checks;
- prior audit findings and generated reports;
- requested output language;
- GitHub issue or pull-request context;
- verification output needed to complete the task.
Enji Guard and the Enji Fleet agents orchestrator should send only the context needed for the selected task.
What Enji Does Not Do
Enji Guard does not:
- train Enji-owned foundation models on Customer Content;
- sell Customer Content;
- use Customer Content to build a competing customer-code model;
- intentionally send customer code to consumer AI products or routes that use customer code for model training.
Third-Party Provider Commitments
The intended hosted-product policy is:
- use the approved third-party providers listed below;
- rely on each provider’s public data-use and retention terms for the selected production route;
- document routes that retain limited abuse-monitoring logs, application state, hosted share content, or other technical data;
- avoid saying “no retention”, “zero retention”, or “privacy mode” globally unless that specific provider route is covered by a separate written commitment or product configuration.
Approved AI And Coding Tool Providers
This page lists providers and product routes, not individual model names.
| Provider | Product/route used by Enji Guard | Data sent | Public no-training status | Public retention status | Notes |
|---|---|---|---|---|---|
| OpenAI | OpenAI API or OpenAI business/developer services | Task context and selected customer content | OpenAI states API/business data is not used for training by default unless the customer opts in | OpenAI API abuse-monitoring logs may retain customer content for up to 30 days by default; Zero Data Retention or Modified Abuse Monitoring requires approval | Hosted Enji Guard should rely on OpenAI’s public route-specific terms unless a separate written commitment applies. |
| Anthropic, PBC | Anthropic API and related commercial API routes | Task context and selected customer content | Anthropic states retained data is not used for model training without express permission | Anthropic offers Zero Data Retention for eligible Claude API routes; other features may retain only the technical data required for the feature | Hosted Enji Guard should rely on Anthropic’s public route-specific terms unless a separate written commitment applies. |
| MOONSHOT AI PTE. LTD. | Kimi OpenPlatform API | Task context and selected customer content | Public Kimi OpenPlatform terms should not be treated as a blanket no-training commitment | Public Kimi OpenPlatform policy says input/account data may be retained while the account is active and as needed for service, security, legal, and business purposes | Use the Singapore-operated Kimi OpenPlatform route. Do not claim no-training or zero-retention for Moonshot unless separate written terms apply. |
| ANOMALY INNOVATIONS, INC. | OpenCode / OpenCode Enterprise / OpenCode Zen, depending on production configuration | Coding-agent context and selected outputs | OpenCode says its open-source/enterprise workflow does not store code or context data | OpenCode open-source/enterprise processing is local or through the selected AI provider; OpenCode Zen publishes provider-specific retention exceptions | Disable hosted share links unless Enji.ai intentionally wants hosted OpenCode share pages. |
| Anysphere, Inc. | Cursor platform, CLI, agents, or related Cursor services if used in production workflows | Coding-agent context and selected outputs | Cursor says Privacy Mode prevents code training by Cursor or third parties | Cursor says Privacy Mode enables zero data retention for model providers, while some feature data, embeddings, metadata, and temporary encrypted cache may be stored | Do not claim Cursor zero-retention for hosted Enji Guard unless the specific route and configuration support it. |
Customer-Controlled Deployment Option
Customers with strict provider, residency, or retention requirements can contact Enji.ai about a separately priced and contracted on-premises or customer-controlled deployment.
In that deployment model, the customer is responsible for choosing, approving, configuring, and monitoring the AI model providers or coding tools connected to their environment. The customer’s selected providers and routes may differ from the hosted Enji Guard provider list.
Provider Reference Links
- OpenAI data controls: https://developers.openai.com/api/docs/guides/your-data
- Anthropic API and data retention: https://platform.claude.com/docs/en/manage-claude/api-and-data-retention
- Kimi OpenPlatform terms: https://platform.kimi.ai/docs/agreement/modeluse
- Kimi OpenPlatform privacy policy: https://moonshot-ai.gitbook.io/moonshot-ai/terms-and-agreement/kimi-openplatform-privacy-policy
- OpenCode terms: https://opencode.ai/legal/terms-of-service
- OpenCode enterprise data handling: https://dev.opencode.ai/docs/enterprise/
- OpenCode Zen privacy notes: https://dev.opencode.ai/docs/zen
- Cursor terms: https://cursor.com/en-US/terms-of-service
- Cursor data use: https://cursor.com/en-US/data-use
Generated Output
AI and agent output can be wrong, incomplete, unsafe, or outdated.
You are responsible for reviewing, testing, and approving reports, recommendations, GitHub issues, pull requests, code changes, and any other output before relying on it.
Customer Controls
Depending on your product configuration, you can reduce AI processing by:
- not connecting a repository;
- not running audits or improvement jobs;
- disabling schedules;
- revoking GitHub App access;
- deleting reports or projects where supported;
- contacting [email protected] for privacy or deletion requests.