Glossary
Data exfiltration
Moving private data, secrets or another tenant’s data to a place it should not go, such as an external email address, URL or public repository.
Checks run on our GPUs in Switzerland. Request payloads are not stored.Security
What Data exfiltration means
Data exfiltration is the unauthorised transfer of data out of the system that should hold it. For agents it rarely looks like a dramatic breach. It looks like an ordinary tool call: an email to an unfamiliar address with a file attached, an HTTP request whose URL carries an API key, a commit that includes a .env file, or a support reply that quotes another customer’s record.
Exfiltration often follows indirect prompt injection and is the final leg of the lethal trifecta: the agent can read private data, reads untrusted instructions, and has a way to send data out.
MCP Guard’s exfiltration head scores whether an action sends private data, secrets or another tenant’s data where it should not go. Deterministic controls help too: egress allowlists, scoped credentials and redacting secrets before they ever reach the model.
Related terms
How Data exfiltration connects to the rest of the vocabulary.