Technology ❯ Artificial Intelligence ❯ AI Security ❯ Model Vulnerabilities
A study found the app can be prompted to voice its internal safety rules, intensifying concerns over fragile moderation.
