Discover AI you didn't know about

Discovery finds AI in use that isn't in your inventory yet, and suggests a starting risk tier for each.

Sources

• Network logs (start here): upload a web proxy, DNS or firewall export. No access is needed. V77 Trust matches host names against 100+ known AI services and keeps only counts, never user names or IP addresses. Files up to 50 MB; for bigger exports, upload a few days at a time.
• Microsoft 365 & Azure: finds AI apps connected to staff accounts, Microsoft 365 Copilot licences, and Azure OpenAI, AI Foundry and ML deployments. Needs a read-only app registration.
• GitHub: finds repositories that use AI libraries (Python and JavaScript/TypeScript projects). Needs a read-only token.
• OpenAI and Anthropic organisations: list projects or workspaces and the models they used in the last 30 days. These need an admin key; such keys can't be limited to read-only, so create a separate key just for V77 Trust. V77 Trust only reads and stores the key encrypted.
• AWS: Bedrock agents, with the model each one runs on and its tools (action groups). A tool that can move money, run code, send messages or change data is flagged. Also finds fine-tuned Bedrock models and SageMaker endpoints. Uses a read-only IAM user with five List/Get permissions.
• Google Cloud: agents on Vertex AI Agent Engine and the models served from Vertex AI endpoints. Uses a service account with the Vertex AI Viewer role.
• Okta: AI apps staff sign in to with company single sign-on. Uses an API token from a Read-Only Administrator.

When you accept a discovered agent, it is added to your inventory and to the agent register with the tools discovery found, so the agent checks and live monitoring apply straight away.

Review what's found

Each result appears under To review with why it was suggested. For each one:

• Add to inventory creates an AI system you can then classify.
• Ignore hides it (you can undo this).
• The … menu lets you link it to a system you already have, or mark it as not approved.

Connected sources re-scan every 24 hours. New finds trigger a notification for AI owners.