Discover
- List every AI tool in use. Ask teams directly, and check browser history, DNS logs or expense reports for AI subscriptions.
- Find AI features hiding in existing apps. Many SaaS tools (email, CRM, note-taking, meeting recorders) added AI features that send data to third parties.
- Identify your sensitive data. Where do customer records, patient data, payment info, contracts and credentials live?
Decide
- Choose approved tools. Pick one or two business AI plans with data protection terms and make them easy to use.
- Write a one-page AI policy. Approved tools, data that must never be entered, and who to ask. Get staff acknowledgement.
- Check data retention settings on approved tools (training opt-outs, chat history, admin controls).
Enforce
- Inspect what's pasted and uploaded into AI sites and block sensitive content, rather than blocking AI entirely.
- Match real data, not just patterns, to avoid false alarms that train people to click through warnings.
- Cover remote and travelling staff, not just the office network.
- Broker AI agent & MCP credentials: if developers connect Claude Code, Cursor, or Copilot to GitHub, Jira, Linear, Stripe, or internal tools over MCP, use 1-click presets on a central MCP gateway with tool approvals and block direct unbrokered MCP traffic from laptops.
Prove
- Keep evidence: signed policy, control screenshots and a monthly export of blocked events, ready for auditors, clients and insurers.
Short on time? Arx Cloud covers the enforce and prove steps for small teams in one app, from 5 users with a 30-day free trial. See the full guide to stopping ChatGPT data leaks.