Test AI agent for rule compliance and safety

Offering · from RUB 4,900 · done in 6 days · AI agents & automation · Working languages: Russian, English · machine translation from Russian I will create a set of safe test dialogues to evaluate whether your AI agent follows given restrictions, handles inappropriate requests, and escalates complex cases to a human. Result: up to 30 synthetic tests, a deviation table, and suggestions for fixes. I do not offer real penetration testing or a guarantee that the model cannot be hacked.