179 questions
No questions match those filters.
← All questions
EvaluationHard· Security & Safety· Alignment
What is the difference between safety capability and sa...
This is one of the questions in the full AI/ML interview bank. Pro unlocks all 1789 questions; Premium includes the same bank plus the highest daily Practice limit.
See plansSafety capability refers to the knowledge encoded in the model’s weights that could potentially be used for harm. Safety propensity refers to the model’s tendency to withhold that information as a result of alignment training. For closed-weight APIs, only the propensity is directly attackable, whereas for open-weight models, the capability itself is accessible and must be audited.