Project files and code
All source code, spreadsheets, text documents, and configuration files remain in their original directories on your computer. The Harness only reads files in folders you explicitly open.
Data architecture
Privacy and security in TritonAI Harness
TritonAI Harness uses a local-first architecture. Your project files, transcripts, and credentials stay on your computer. This statement explains what data stays local, what passes through the gateway, and how UC policies apply.
Unlike centralized web chat interfaces, the Harness does not retain your work on campus servers. You keep custody of your local files and project history.
Local storage
What remains on your computer
The Harness stores operational data locally within your user account rather than on central servers.
Transcripts and logs
Task transcripts, agent execution steps, and interaction histories are written directly to your local application directory. They are not stored on central campus servers.
Local memory vaults
Persistent notes, context summaries, and reference instructions are saved as local Markdown files on your workstation. No cloud indexing or external training takes place.
Network transit
What crosses the campus gateway
Communication between your workstation and TritonAI infrastructure is limited to active model requests and operational telemetry.
Active prompt context & zero silent fallback
When an agent turn executes, only the prompt text and specific file snippets needed for that turn travel across encrypted HTTPS to the TritonAI Gateway.
Campus on-premises models hosted on campus infrastructure process requests with zero data retention and no model training. Your prompts are discarded after generating the response.
The Harness maintains an empty cloud fallback list by default. If an on-premises model is temporarily unavailable, the error is surfaced immediately. Private campus context is never silently redirected to external cloud providers.
Usage and billing telemetry
The Gateway logs request timestamps, model identifiers, token counts, and account identifiers to monitor platform availability, manage quotas, and apply recharge billing where applicable.
The Gateway does not record the text content of your local files, shell command outputs, or conversation transcripts in its telemetry logs.
Credential boundaries
Plugin security and credential isolation
TritonAI Harness uses host-managed authentication to isolate credentials from language models.
Host-managed tokens
OAuth tokens for GitHub, Google Workspace, and Microsoft 365 reside in the local host secret store. Tokens are never passed into model prompt context.
Application API execution
The Harness application process makes authorized API calls directly. The AI agent only proposes structured parameters for each action.
Draft-first safeguards
Write operations for email and calendar default to draft mode. External changes require explicit human confirmation before sending or publishing.
Immediate tool revocation
Each plugin can be toggled on or off in Settings. Disabling access immediately revokes tools before the next agent turn.
Policy compliance
Approved UC protection levels
Use of the Harness must align with University of California Electronic Information Security Policy (IS-3).
Approved for P1 through P3
TritonAI Harness is approved for UC information classified as Protection Level 1 (Public), Protection Level 2 (Internal), and Protection Level 3 (Sensitive) within approved campus setups.
Because project files and transcripts reside on your local computer, your workstation must meet UC San Diego minimum security standards, including active disk encryption (FileVault on macOS, BitLocker on Windows).
Prohibited for P4 data
Protection Level 4 (P4) data is strictly prohibited in TritonAI Harness. This includes electronic protected health information subject to HIPAA, credit card numbers subject to PCI DSS, and high-consequence regulatory data.
For health system patient-care workflows, consult the approved clinical AI services available through the UC San Diego Health Pulse portal.
User control
Data deletion and access revocation
Because session data is stored locally, you maintain complete control over data removal.
You can delete past task transcripts, remove cached files, or clear memory entries at any time by removing them from your local directory or clearing history within the application interface. Deletion is instantaneous and permanent.
To revoke Microsoft 365 access, use the Disconnect button in Settings under Plugins, or revoke permissions directly in your Microsoft 365 account security settings.
Governance contact
Questions about privacy and data use
For questions regarding TritonAI data governance, security assessments, or approved use cases, contact the service team.