Privacy and Governance Fundamentals
Who this is for
Organization Owners evaluating AI Pilot for production use, compliance officers, and anyone who needs to understand how AI Pilot handles data before enabling it.
What you will complete
Understand what data AI Pilot can access, what it cannot access, how data is handled, and what governance controls are available to your organization.
What data AI Pilot can access
AI Pilot reads the following data to answer questions and propose actions:
- Server metrics — CPU, memory, disk usage, network throughput (aggregated, not raw packets)
- Alert history — which alerts fired, when, and on which server
- Server and site listings — names, statuses, provider, region
- Deployment history — which deployments ran, when, and their results
- Backup history — which backups exist, their size and status
- Service and log output — when File and Log Access is enabled, the AI can read logs via SSH
AI Pilot reads this data only when you start a conversation or ask a question. It does not continuously monitor or process your data in the background.
What data AI Pilot cannot access
These are hard limits that cannot be changed by configuration:
- Environment variable values — the AI can see that a variable named DATABASE_URL exists, but it cannot read the value
- Cloud API keys and credentials — stored in the platform encrypted; never exposed to the AI
- Database contents — the AI cannot query your databases or read table data
- User passwords and session tokens — never accessible to the AI
- Private file contents (unless you explicitly share a file path in chat and File Access is enabled)
How AI Pilot handles your prompts
When you send a message to AI Pilot:
- Your message and the relevant infrastructure context (server metrics, alert state) are sent to the AI provider you configured (Anthropic, OpenAI, Google, or Ollama).
- The AI provider processes the request and returns a response.
- The response is displayed in your chat.
- The conversation is logged in the audit trail.
For Ollama (local): Your prompts and infrastructure context are processed entirely on your own infrastructure. No data leaves your network.
For cloud providers (Anthropic, OpenAI, Google): Your prompts and context are sent to that provider's API. Subject to that provider's data processing terms. CloudAIPilot does not store prompts long-term beyond the audit trail.
The learning corpus and opt-out
After conversations, CloudAIPilot extracts scrubbed lessons to improve AI Pilot for all users. Before extraction:
- Hostnames, IP addresses, and server names are replaced with placeholders.
- Environment variable names and values are removed.
- File paths are anonymized.
- No raw user messages or code are stored in the learning corpus.
To opt your organization out entirely: go to Settings → AI Agent → Agent Controls → Opt out of AI Pilot learning corpus and enable the toggle.
Governance controls available to your organization
| Control | Where to find it | What it does |
|---|---|---|
| Master AI switch | Settings → AI Agent → Agent Controls | Disable AI Pilot entirely |
| Write Actions toggle | Settings → AI Agent → Agent Controls | Prevent any write operations |
| Production Protection | Settings → AI Agent → Agent Controls | Block writes on production servers |
| Per-server access | Settings → AI Agent → Per-Server Access | Restrict AI access per server |
| Fine-tune Write Actions | Settings → AI Agent → Agent Controls | Disable specific operation types |
| Auto Memory toggle | Settings → AI Agent → Agent Controls | Stop the AI from storing new memories |
| Learning corpus opt-out | Settings → AI Agent → Agent Controls | Stop lesson extraction from your org |
| Audit trail | Settings → Audit Log | Review all AI actions permanently |
Roles and AI Pilot access
| Role | Can chat with AI? | Can approve write actions? | Can change AI settings? |
|---|---|---|---|
| Viewer | No | No | No |
| Member | Yes (read + limited write) | Yes (within Member scope) | No |
| Admin | Yes | Yes | Yes |
| Owner | Yes | Yes | Yes |
What success looks like
- Your organization's governance requirements are met using the controls listed above.
- The audit trail provides a complete record of AI activity for compliance review.
- Sensitive values (credentials, database contents) never appear in AI Pilot chat or audit logs.
Common questions
"Can AI Pilot see my SSH private keys?" No. SSH keys are stored encrypted and are never passed to the AI. The AI uses platform-managed SSH access to run commands, but it does not have access to the underlying key material.
"If I use OpenAI as my provider, does OpenAI store my infrastructure data?" Your prompts and context are sent to OpenAI for processing under their API terms. Check OpenAI's data processing agreement for enterprise data retention policies. For maximum data sovereignty, use Ollama (local).
"Can the AI read my environment variables?" The AI can see environment variable names. It cannot read values. This is a hard limit.
"Is there a way to completely air-gap AI Pilot from external APIs?" Yes — select Ollama (Local) as your AI provider. All AI processing happens on your own infrastructure.