The short answer
The main risks are confident wrong answers, one customer’s data reaching another, and users trying to trick the assistant into ignoring its rules (prompt injection).
Practical defences: answer from retrieved facts, separate each business’s data, treat all user text as untrusted, and require approval for real-world actions.
What GorZesk does about it
GorZesk keys all data to a business, tests for cross-tenant access and prompt-injection attempts in an automated evaluation suite, and requires approval before consequential actions.
Limits to know
- No AI system is risk-free. Pinevon does not claim a security certification; see the Trust pages for what is and is not in place.
Common questions
Does Pinevon hold a security certification?
No. The Trust page lists what is not claimed as clearly as what is.
Related topics
Last updated 2026-09-25. Something unclear or wrong? Tell us.