Glossary: Application and Credential Attacks, and AI Risks
guardrails
Also written: guardrail, safety rules
The safety limits built into an AI, such as "don't write malware" or "don't give out personal data". Jailbreaking tries to get around them.
Glossary: Application and Credential Attacks, and AI Risks
Also written: guardrail, safety rules
The safety limits built into an AI, such as "don't write malware" or "don't give out personal data". Jailbreaking tries to get around them.